A short-term wind power prediction method considering gust probability statistics and correction

Abstract

The application provides a short-term wind power prediction method considering gust probability statistics and correction, which is based on existing historical data of a wind farm, and the method statistically predicts the probability of gust occurrence at each time in the previous quarter, the probability distribution of gust duration, and the probability distribution of wind speed offset coefficient in the gust period. Random numbers are generated at each time of the wind speed prediction sequence to simulate the probability of gust occurrence, and the wind speed prediction sequence is corrected at the time when the gust occurs. The historical power generation data, meteorological data and corrected meteorological prediction data before the prediction starting time are input into a neural network to generate a predicted wind power sequence. The application considers the wind energy resource conditions of the wind farm in the correction of the meteorological prediction by gust probability statistics and correction, thereby improving the data quality of the wind speed prediction, reducing the occurrence of large wind power prediction deviation caused by inaccurate wind speed prediction, and improving the accuracy of the short-term power prediction of the wind farm.

Description

Technical Field

[0001] This invention relates to the field of wind power generation, and mainly to a short-term wind power prediction method that takes into account gust probability statistics and corrections. Background Technology

[0002] With the increasing consumption of Earth's resources, energy issues are becoming increasingly prominent. Vigorously developing renewable energy technologies such as photovoltaic and wind power generation is a crucial means of alleviating these problems. With the large-scale construction of wind farms of varying sizes across the country, wind power, a distributed energy source characterized by randomness, volatility, and intermittency, will occupy an increasingly significant proportion of my country's power grid. Wind power generation largely depends on uncertain and uncontrollable meteorological factors, such as wind speed, wind direction, and temperature. Changes in wind direction and speed can cause irregular fluctuations in power generation, posing a significant challenge to the existing power grid system. Accurate wind power forecasting can reduce the impact of wind power uncertainty on the power grid, maintain power supply reliability and quality, help power system dispatching departments adjust dispatch plans in a timely manner, reduce wind curtailment, ensure power system stability, and thus increase the adoption of renewable energy generation systems. Furthermore, accurate wind power forecasting can also assist wind farms in bidding for electricity in the market, laying the foundation for my country's power market reform.

[0003] Current weather forecasts are inaccurate in predicting gusts, leading to significant discrepancies between short-term predicted and actual power outputs for wind farms during gust periods. A review of existing patent literature reveals related wind power prediction methods, such as CN202310259340.5, which utilize historical turbine operating data and historical meteorological data to construct time-segmented prediction models. However, when gusts occur, the future wind speed differs significantly from historical trends, greatly impacting the model's predictive accuracy. Summary of the Invention

[0004] The purpose of this invention is to provide a short-term wind power prediction method that takes into account gust probability statistics and corrections. For each prediction, the gust probability statistics for different time periods in the past quarter are obtained, thereby effectively correcting the forecast wind speed sequence. By integrating historical data and meteorological forecast data corrected for gust probability, the occurrence of large errors in wind power prediction caused by gust phenomena is reduced, and the accuracy of short-term wind power prediction is improved.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A short-term wind power prediction method considering gust probability statistics and corrections, the method comprising the following steps:

[0007] (1) Determine the gust occurrence threshold, and determine the time when gusts occur and end in different time periods of the previous quarter based on the gust occurrence threshold;

[0008] (2) Calculate the duration of each gust and the deviation coefficient of the wind speed at each moment during each gust period relative to the average wind speed during the same period in the previous quarter when there were no gusts.

[0009] (3) Calculate the probability of gusts occurring at each time of day in the previous quarter, and calculate the probability distribution of the duration of gusts and the wind speed deviation coefficient during the gust period in the three time periods.

[0010] (4) Generate random numbers for each moment of the wind speed forecast sequence to simulate the probability of gusts occurring and determine whether gusts occur at each moment.

[0011] (5) Correct the wind speed forecast sequence: If gusts occur during a certain period, then correct it;

[0012] (6) Input the historical power generation data of the wind farm before the prediction start time, meteorological data, and corrected meteorological forecast data into the neural network to generate the predicted wind power sequence.

[0013] At least one embodiment of the present invention discloses a short-term wind power prediction method that considers gust probability statistics and corrections. In step (1), the specific process of determining the gust occurrence threshold and determining the occurrence and end times of gusts in different time periods of the previous quarter based on the gust occurrence threshold is as follows:

[0014] (1.1) Considering that historical wind field data is often time-resolutiond at 15 minutes, a time window of 4 hours, from t-16 to t, is set for time t, and the average wind speed within the window is calculated. and peak wind speed t-16～t :

[0015]

[0016] peak t-16～t =max(wspd) t-16 wspd t-15 ,…wspd t )

[0017] in, It is t i Wind speed at any given time;

[0018] (1.2) Settings The threshold for gust occurrence within this time window, if And wspd t It is the peak value within that time window (i.e., wspd).t =peak t-16～t If the wind gust occurs at time t, then it is assumed that a gust of wind occurs at time t.

[0019] (1.3) Repeat steps (1.1) and (1.2) for all times, determine whether there is a gust of wind at that time, and count the times when gusts of wind occur in the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:00;

[0020] (1.4) At time t when the gust occurs, set a time window of 2 hours from t to t+8, and calculate the average wind speed within the window. Valley wind speed t～t+8 :

[0021]

[0022] valley t～t+8 =min(wspd) t wspd t+1 ,…wspd t+8 )

[0023] (1.5) Settings This is the threshold for the end of gusts within this time window. And wspd t It is the valley value within that time window (i.e., wspd) t =valley t～t+8 If the gust ends at time t, then it is considered to have ended at time t.

[0024] (1.6) Repeat (1.4) and (1.5) for all gust times to determine whether the gust ends at that time, and count the times when the gust ends in the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:00.

[0025] Preferably, the formula for determining the occurrence of gusts Formula for determining the end of a gust of wind The coefficient n can be selected as 0.2.

[0026] At least one embodiment of the present invention discloses a short-term wind power prediction method that considers gust probability statistics and correction. In step (2), the specific process of calculating the duration of each gust and the offset coefficient of the wind speed at each moment of each gust period relative to the average wind speed when there were no gusts in the previous quarter is as follows:

[0027] (2.1) From the wind speed data, the time t of the first occurrence of a gust is determined as the time of the nearest gust onset. begin and the end time t endTwo are determined as a gust period. i Its duration is:

[0028]

[0029] (2.2) For the gust period in (2.1) i The wind speed offset coefficient at each time point is:

[0030]

[0031] in, It is gust i Time period j Average wind speed at each time point during non-gust periods; wspd i Let i be the wind speed at time i;

[0032] The calculation formula is:

[0033]

[0034] Where 8 represents period j The duration is 8 hours;

[0035] t i This is a period when there are no gusts of wind.

[0036] length(gust) is period j The total duration of all gust periods in the middle;

[0037] period j,start It is period j The beginning moment;

[0038] period j,end It is period j The end time;

[0039] (2.3) Repeat steps (2.1) and (2.2) for all gust periods to calculate the duration of all gusts and the offset coefficient of the wind speed at each moment during the gust period relative to the average wind speed when there are no gusts during that period.

[0040] At least one embodiment of this disclosure provides a short-term wind power prediction method that considers gust probability statistics and corrections. In step (3), the specific process of statistically analyzing the probability of gusts occurring at each time of day in the previous quarter and calculating the probability distribution of gust duration and gust wind speed deviation coefficient for three time periods is as follows:

[0041] (3.1) Calculate the probability of gusts occurring at different times of day in the previous quarter, and calculate the probability of gusts for each day.i The probability of gusts occurring is:

[0042]

[0043] Among them, t i,gust It means t per day i The value is 1 if a gust of wind occurs, and 0 otherwise.

[0044] number(t i,gust ) indicates the daily t in this quarter i The number of times gusts occur;

[0045] totalnumber(t i ) indicates the quarter t i The total number of days in the quarter;

[0046] (3.2) Based on the duration of each gust calculated in step (2) and the offset coefficient of the wind speed at each moment during each gust period relative to the average wind speed when there are no gusts during that period, the probability distribution of the duration of gusts is statistically analyzed for the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:00 respectively: And the probability distribution of the wind speed shift coefficient during gust periods:

[0047] At least one embodiment of this disclosure provides a short-term wind power prediction method that considers gust probability statistics and corrections. In step (4), random numbers are generated for each moment of the wind speed forecast sequence to simulate the probability of gusts occurring. The specific process for determining whether gusts occur at each moment is as follows:

[0048] (4.1) To simulate the probability of gusts occurring at each moment, the Monte Carlo simulation approach is adopted: using random sampling to approximate the behavior of a complex system. A random number sequence r = {r1, r2, r3, ... r...} is generated from the wind speed forecast sequence of length k. i ,…r k}, where any random number r i All samples are randomly drawn from a uniform distribution in [0,1], i.e., r i ~U(0,1), and r1,r2,r3,…,r k The numbers are independent of each other. The generated random number sequence is used to simulate the probability of gusts occurring, r. i This represents the probability of a gust occurring at a given time point, thus simulating the probability of a gust occurring.

[0049] (4.2) For t i At any time, if Then determine t iGusts of wind can occur at any time.

[0050] At least one embodiment of the present invention provides a short-term wind power prediction method that considers gust probability statistics and corrections. In step (5), the wind speed forecast sequence is corrected: if a gust occurs during a certain period, the specific process of correction is as follows:

[0051] For gusts i t i At that moment, a gust of wind i Duration (length(gust)) i () is the period to which it belongs. j The expected value of the probability distribution of the duration of gusts:

[0052]

[0053] Where, length k For a certain gust of wind k Duration;

[0054] K represents the period in the previous quarter. j The number of times gusts occurred within a given time period;

[0055] Indicates the period of the previous quarter. j Length in time period k The proportion of all gust durations.

[0056] Gust i Wind speed at each moment in a time period is represented by its corresponding period. j The expected value of the probability distribution of the wind speed offset coefficient is used for correction:

[0057]

[0058] in, For the t where gusts occur i Wind speed value after real-time correction;

[0059] For t i Original wind speed value at any given time;

[0060] M represents the period in the previous quarter. j The total number of times during which gusts occurred within a given time period;

[0061] offset m For the previous quarter period j Time period gust time t m The wind speed offset coefficient;

[0062] representing the period of the previous quarter j the offset in the time period m the proportion of all offset coefficients.

[0063] Preferably, for time t2, if it is already in the gust period starting from time t1 (t1<t2) and it is determined that a gust occurs, the gust is considered to continue to and the wind speed of the extended gust period is also subjected to the corresponding correction in step (5).

[0064] At least one embodiment disclosed in the present invention provides a short-term wind power prediction method considering gust probability statistics and correction. In step (6), historical power generation data and meteorological data of the wind farm are input into a gated recurrent unit network. During the training process, the gated recurrent unit network learns the relationship between meteorological data and wind power from a large amount of historical power generation data and meteorological data of the wind farm, and updates network parameters according to the prediction error to complete the training of the gated recurrent unit network. Then, the wind farm historical data of a certain length before the prediction time and the corrected meteorological data are input into the gated recurrent unit network, to infer and output the wind power prediction sequence.

[0065] The present invention has the following beneficial effects:

[0066] (1) On the basis that the existing short-term wind power prediction methods rely on meteorological forecast (especially wind speed forecast) data, the present invention counts the gust probability statistical results of different time periods in the previous quarter for each prediction start time point, so as to effectively correct the forecast wind speed sequence and improve the accuracy of the wind speed forecast data.

[0067] (2) The present invention integrates historical data and meteorological forecast data corrected by gust probability, comprehensively considers the wind energy resource conditions of the wind farm contained in the historical data and the higher-precision wind speed forecast, reduces the occurrence of extremely large errors in wind power prediction caused by gust phenomena, and improves the short-term wind power prediction accuracy of the wind farm. Brief Description of the Drawings

[0068] Figure 1 shows the flow chart of the short-term wind power prediction model of the present invention.

[0069] Figure 2 shows the specific flow chart of using random numbers to simulate the gust occurrence probability and judge whether a gust occurs in the present invention. Detailed Description of the Embodiments

[0070] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that the embodiments are only specific illustrations of the invention and should not be regarded as limitations on the invention. The purpose of the embodiments is to enable those skilled in the art to better understand and reproduce the technical solution of the present invention. The scope of protection of the present invention should still be determined by the scope defined in the claims.

[0071] like Figure 1 As shown in the figure, this embodiment provides a short-term wind power prediction method that considers gust probability statistics and corrections. The method is as follows:

[0072] S1, determine the gust occurrence threshold, and determine the time when gusts occur and end in different time periods of the previous quarter based on the gust occurrence threshold;

[0073] The specific process is as follows:

[0074] S11. Considering that historical wind field data is often time-resolutiond with 15-minute intervals, a time window of 4 hours, from t-16 to t, is set for time t, and the average wind speed within the window is calculated. and peak wind speed t-16～t :

[0075]

[0076] peak t-16～t =max(wspd) t-16 wspd t-15 ,…wspd t )

[0077] in, It is t i The wind speed value at any given time.

[0078] S12, Settings The threshold for gust occurrence within this time window, if And wspd t It is the peak value within that time window (i.e., wspd). t =peak t-16～t If ), then it is assumed that a gust of wind occurs at time t.

[0079] S13. Repeat steps S11 and S12 for all times to determine whether a gust of wind occurs at that time. Count the times when gusts of wind occur in the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:00.

[0080] S14. At time t when the gust occurs, set a time window of 2 hours from t to t+8, and calculate the average wind speed within the window. Valley wind speedt～t+8 :

[0081]

[0082] valley t～t+8 =min(wspd) t wspd t+1 ,…wspd t+8 )

[0083] S15, Settings This is the threshold for the end of gusts within this time window. And wspd t It is the valley value within that time window (i.e., wspd) t =valley t～t+8 If the gust ends at time t, then it is considered to have ended at time t.

[0084] S16. Repeat steps S14 and S15 for all gust times to determine whether the gust ends at that time. Count the times when the gust ends in the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:00.

[0085] Preferably, the formula for determining the occurrence of gusts Formula for determining the end of a gust of wind The coefficient n can be selected as 0.2.

[0086] S2, calculate the duration of each gust, and the deviation coefficient of the wind speed at each moment during each gust period relative to the average wind speed during the same period without gusts in the previous quarter.

[0087] The specific process is as follows:

[0088] S21, from the first determination of the occurrence of a gust in the wind speed data, the time t of the start of the nearest gust. begin and the end time t end Two are determined as a gust period. i Its duration is:

[0089]

[0090] S22, for the gust period in S21 i The wind speed offset coefficient at each time point is:

[0091]

[0092] in, It is gust i Time period jAverage wind speed at each time point during non-gust periods; wspd i Let i be the wind speed at time i;

[0093]

[0094] Where 8 represents period j The duration is 8 hours;

[0095] t i This is a period when there are no gusts of wind.

[0096] length(gust) is a period j The total duration of all gust periods in the middle;

[0097] period j,start It is period j The beginning moment;

[0098] period j,end It is period j The end time;

[0099] S23. Repeat S21 and S22 for all gust periods to calculate the duration of all gusts and the offset coefficient of the wind speed at each moment during a gust period relative to the average wind speed when there are no gusts during that period.

[0100] S3, calculate the probability of gusts occurring at each time of day in the previous quarter, and calculate the probability distribution of gust duration and gust wind speed deviation coefficient for three time periods respectively.

[0101] The specific process is as follows:

[0102] S31, Calculate the probability of gusts occurring at different times of day in the previous quarter, t per day. i The probability of gusts occurring is:

[0103]

[0104] Among them, t i,gust It means t per day i The value is 1 if a gust of wind occurs, and 0 otherwise.

[0105] number(t i,gust ) indicates the daily t in this quarter i The number of times gusts occur;

[0106] totalnumber(t i ) indicates the quarter t i The total number of days in the quarter;

[0107] S32, based on the duration of each gust calculated in S2 and the offset coefficient of the wind speed at each moment during each gust period relative to the average wind speed when there are no gusts during that period, the probability distribution of the duration of gusts is statistically analyzed for the three time periods: 4:00~12:00, 12:00~20:00, and 20:00~4:00. And the probability distribution of the wind speed shift coefficient during gust periods:

[0108] S4 generates random numbers for each moment of the wind speed forecast sequence to simulate the probability of gusts occurring and determines whether gusts occur at each moment.

[0109] like Figure 2 As shown, the process of generating random numbers for each moment in the wind speed forecast sequence to simulate the probability of gusts occurring, and determining whether gusts occur at each moment, is as follows:

[0110] S41. To simulate the probability of gusts occurring at each moment, the Monte Carlo simulation approach is adopted: using random sampling to approximate the behavior of a complex system. A random number sequence r = {r1, r2, r3, ... r...} is generated from the wind speed forecast sequence of length k. i ,…r k}, where any random number r i All samples are randomly drawn from a uniform distribution in [0,1], i.e., r i ~U(0,1), and r1,r2,r3,…,r k The numbers are independent of each other. The generated random number sequence is used to simulate the probability of gusts occurring, r. i This represents the probability of a gust occurring at a given time point, thus simulating the probability of a gust occurring.

[0111] S42, for t i At any time, if Then determine t i Gusts of wind can occur at any time.

[0112] S5, corrects the wind speed forecast sequence: if gusts occur during a certain period, it is corrected;

[0113] For gusts i t i At that moment, a gust of wind i Duration (length(gust)) i () is the period to which it belongs. j The expected value of the probability distribution of the duration of gusts:

[0114]

[0115] where length k is the duration of a certain gust k ;

[0116] K is the number of gusts occurring in the period j time period in the previous quarter;

[0117] represents the proportion of length j in the period k time period of the previous quarter accounting for the total duration of all gusts.

[0118] The wind speed at each moment in the gust i period is corrected using the expected value of the probability distribution of the wind speed offset coefficient of the period j to which it belongs:

[0119]

[0120] where, is the corrected wind speed value at moment t when the gust occurs i ;

[0121] is the original wind speed value at moment t i ;

[0122] M is the total number of moments at which gusts occur in the period j time period in the previous quarter;

[0123] offset m is the wind speed offset coefficient at the gust moment t in the period j time period in the previous quarter m ;

[0124] represents the proportion of offset in the period j time period of the previous quarter m accounting for all offset coefficients.

[0125] Preferably, for moment t2, if it is already in the gust period starting from moment t1 (t1<t2) and it is determined that a gust occurs at t2, the gust is considered to continue until and the wind speed of the extended gust period is also subjected to the corresponding correction in S5.

[0126] S6: input the historical power generation data of the wind field before the prediction start moment, the meteorological data, and the corrected meteorological forecast data into a neural network to generate a predicted wind power sequence.

[0127] Historical wind farm power generation data and meteorological data are input into a gated recurrent unit (GRU) network. During training, the GRU learns the relationship between meteorological data and wind power from a large amount of historical wind farm power generation data and meteorological data, and updates its network parameters based on prediction errors to complete the training of the GRU. Then, historical wind farm data for a certain period before the prediction time and the corrected meteorological data are input into the GRU to infer and output a wind power prediction sequence.

[0128] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

Claims

1. A short-term wind power prediction method considering gust probability statistics and corrections, characterized in that, The method is as follows: (1) Determine the gust occurrence threshold, and determine the time of gust occurrence and end in different time periods of the previous quarter based on the gust occurrence threshold; (2) Calculate the duration of each gust and the offset coefficient of the wind speed at each moment during each gust period relative to the average wind speed during the same period without gusts in the previous quarter. (3) Calculate the probability of gusts occurring at each time of day in the previous quarter, and calculate the probability distribution of the duration of gusts and the wind speed deviation coefficient during the gust period for the three time periods respectively. (4) Generate random numbers for each moment of the wind speed forecast sequence to simulate the probability of gusts occurring and determine whether gusts occur at each moment; (5) Correct the wind speed forecast sequence: If If gusts occur during a given period, adjustments will be made accordingly. (6) Input the historical power generation data of the wind farm, meteorological data and the corrected meteorological forecast data before the start of the prediction into the neural network to generate the predicted wind power sequence; Step (4) specifically includes the following steps: (4.1) The Monte Carlo simulation method is used to simulate the probability of gusts occurring at each moment. The specific method is as follows: for a length of... Generate a random number sequence from the wind speed forecast sequence. any random number They are all randomly drawn from a uniform distribution of [0,1], that is ,and The random number sequences generated are independent of each other; they are used to simulate the probability of gusts occurring at each time point in the wind speed forecast sequence. This represents the probability of a gust occurring at a given time point, thus simulating the probability of a gust occurring. (4.2) For At any time, if Then determine Gusts of wind can occur at any time; Step (5) specifically includes the following steps: For gusts of wind of At that moment, a gust of wind Duration It is the time period it belongs to. The expected value of the probability distribution of the duration of gusts: in, For a certain gust of wind Duration; For the previous quarter The number of times gusts occurred within a given time period; Indicates the previous quarter During the time period The proportion of all gust durations; gust Wind speed at each point in time is determined by the time period to which it belongs. The expected value of the probability distribution of the wind speed offset coefficient is used for correction: in, For the occurrence of gusts Wind speed value after real-time correction; for Original wind speed value at any given time; For the previous quarter The total number of times during which gusts occurred within a given time period; For the previous quarter Time period gust moment The wind speed offset coefficient; Indicates the previous quarter During the time period The proportion of all offset coefficients.

2. The short-term wind power prediction method considering gust probability statistics and corrections according to claim 1, characterized in that, Step (1) specifically includes the following steps: (1.1) For At any time, set one Calculate the average wind speed within a 4-hour time window. and peak wind speed : in, yes Wind speed at any given time; (1.2) Settings The threshold for gust occurrence within this time window, if ,and If it is the peak value within that time window, then it is considered... Gusts of wind occur frequently; (1.3) Repeat steps (1.1) and (1.2) for all times, determine whether there is a gust of wind at that time, and count the times when gusts of wind occur in the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:00; (1.4) The occurrence of gusts At any time, set one Calculate the average wind speed within a 2-hour time window. and wind speed valley value : (1.5) Settings This is the threshold for the end of gusts within this time window. ,and If it is the lowest value within that time window, then it is considered... The gust of wind ended at that moment; (1.6) Repeat (1.4) and (1.5) for all gust times, determine whether the gust ends at that time, and count the times when the gust ends in the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:

00.

3. The short-term wind power prediction method considering gust probability statistics and corrections according to claim 1, characterized in that, Step (2) specifically includes the following steps: (2.1) Determine the time of the first occurrence of a gust from the wind speed data, and the time of the start of the nearest gust. and end time Two are determined as a gust period. Its duration is: (2.2) Regarding the gust periods in (2.1) The wind speed offset coefficient at each time point is: in, yes Time period The average wind speed at each time point during non-gust periods; for Wind speed at any given moment; The calculation formula is: Where 8 represents The duration is 8 hours; This is a period when there are no gusts of wind. for The total duration of all gust periods in the middle; yes The beginning moment; yes The end time; (2.3) Repeat (2.1) and (2.2) for all gust periods to calculate the duration of all gusts and the offset coefficient of the wind speed at each moment of the gust period relative to the average wind speed when there are no gusts in that period.

4. The short-term wind power prediction method considering gust probability statistics and corrections according to claim 1, characterized in that, Step (3) specifically includes the following steps: (3.1) Calculate the probability of gusts occurring at different times of day in the previous quarter. The probability of gusts occurring is: in, Indicates daily The value is 1 if a gust of wind occurs, and 0 otherwise. This indicates the daily situation during this quarter. The number of times gusts occur; This indicates the quarter The total number of days in the quarter; (3.2) Based on the duration of each gust calculated in step (2) and the offset coefficient of the wind speed at each moment during each gust period relative to the average wind speed when there are no gusts during that period, the probability distribution of the duration of gusts is calculated for the three time periods of 4:00~12:00, 12:00~20:00, and 20:00~4:00 respectively: , , And the probability distribution of the wind speed shift coefficient during gust periods: , , .

5. The short-term wind power prediction method considering gust probability statistics and corrections according to claim 1, characterized in that, In step (6), historical wind farm power generation data and meteorological data are input into the gated cyclic unit network. The gated cyclic unit network learns the relationship between meteorological data and wind power, and updates the network parameters according to the prediction error to complete the training of the gated cyclic unit network. Then, historical wind farm data of a certain length before the prediction time and the corrected meteorological data are input into the gated cyclic unit network to infer and output the wind power prediction sequence.

Images