A method for generating wind power output simulation data according to historical data

Abstract

The application relates to a method for generating wind power output simulation data according to historical data, which comprises the following steps: dividing historical wind power output data into several seasons through different date segmentation modes; calculating the expected value of wind power output of each season formed by each date segmentation mode, selecting a date segmentation mode with the maximum expected value deviation between seasons, and dividing the historical wind power output data into several seasons through the date segmentation mode; statistically obtaining the wind power output probability distribution in each season and the Markov transition matrix at each time; and simulating the wind power output at each date and each time in each season by using a Metropolis-Hastings algorithm to obtain annual simulation data. The simulation data generated by the scheme can better conform to the probability distribution of local historical data and the output characteristics of local wind power, and is suitable for Monte Carlo simulation calculation and other calculation researches which require a large amount of data.

Description

Technical Field

[0001] This invention belongs to the field of numerical simulation technology, and specifically relates to a method for generating wind power output simulation data based on historical data. Background Technology

[0002] With the increasing scale of grid-connected wind power capacity, the adverse effects of the randomness, intermittency, and uncertainty of wind power generation systems on the public power grid are gradually becoming apparent. To study the impact of wind power generation uncertainty on the power system and analyze wind power consumption issues, stochastic production simulation technology based on uncertain new energy sources has become one of the main techniques for analyzing new energy consumption.

[0003] Stochastic production simulation technology for uncertain renewable energy requires a large amount of annual wind power output curves as input to study the impact of stochastic variations in wind power output on renewable energy consumption using the Monte Carlo method. Wind power output refers to the amount of electricity generated by a wind farm per unit time. However, most renewable wind farms are currently newly built, and the collected historical wind power output data is relatively limited, which cannot meet the input requirements for stochastic production simulation calculations. Therefore, it is necessary to generate a large amount of simulation data that conforms to local wind power characteristics based on historical wind power output data.

[0004] Current wind power forecasting technologies, such as Chinese invention patents with publication numbers CN115173483A and CN113449471A, tend to rely more on the accuracy of weather-related forecasts. However, the data required for production simulation calculations are more concerned with whether the distribution probability of the data conforms to the historical output characteristics of wind power. Therefore, a simulation data method that pays more attention to the overall distribution of generated data is needed. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method for generating wind power output simulation data based on historical data, thereby solving the problem that the overall distribution of data generated by existing technical solutions does not match the historical output characteristics of local wind power, achieving data that better meets the requirements of production simulation calculations such as Monte Carlo simulation calculations, and thus providing more effective guidance for production.

[0006] According to the technical solution of the present invention, the present invention provides a method for generating wind power output simulation data based on historical data, characterized by comprising the following steps:

[0007] Historical wind power output data is divided into several seasons using different date segmentation methods;

[0008] Calculate the expected wind power output for each season under each date segmentation method, select the date segmentation method with the largest deviation between expected values ​​for each season, and divide the historical wind power output data into several seasons using this date segmentation method.

[0009] The probability distribution of wind power output and the Markov transition matrix at each time point were obtained statistically.

[0010] Using the Metropolis-Hastington algorithm, wind power output was simulated for each date and time in each season to obtain annual simulation data.

[0011] Furthermore, in some embodiments, the method of the present invention includes the following steps:

[0012] Step S1: Collect historical wind power output data for the local area over many years and normalize the data;

[0013] Step S2: Clean the data, remove erroneous data, and replace it with correct data;

[0014] Step S3: Divide the year into segments according to the wind power output characteristics of different seasons; Step S3 further includes:

[0015] Step S3.1: First, consider the local seasonal characteristics and set the seasonal transition period;

[0016] Step S3.2: Select a set of dates during the seasonal transition period to form a set of seasonal day division schemes, dividing the year into Q seasons;

[0017] Step S3.3: Integrate the historical wind power output data from multiple years that belong to the same season into a dataset for that season;

[0018] Step S3.4: Calculate the expected value E for each moment in the dataset for each season using the following formula. q,t and standard deviation σ q,t ;

[0019]

[0020]

[0021] In the formula: q is the season number, q = 1, ..., Q; t is a certain time of day; The wind power output value at time t on a certain day in season q; Let N be the average wind power output at time t for each day in the dataset of season q; N is the total number of days in the dataset of season q.

[0022] Step S3.5: Calculate the expected Euler distance L between seasons using the following formula. E,q Sum of standard deviation and Euler distance L σ,q :

[0023]

[0024]

[0025] In the formula: a and b represent two adjacent seasons respectively. When q≠Q, a=q, b=q+1; when q=Q, a=q, b=1; m is the total number of moments in a day.

[0026] The sum of the expected value Euler distance and the standard deviation Euler distance, L, is calculated using the following formula. sum :

[0027]

[0028] Step S3.6: During the seasonal transition period, select different seasonal day segmentation schemes, and repeat steps S3.2 to S3.5 to calculate the sum L of the expected value Eulerian distance and the standard deviation Eulerian distance corresponding to all possible seasonal day segmentation schemes. sum ;

[0029] Step S3.7: By comparison, the sum L of the expected value Euler distance and the standard deviation Euler distance is obtained. sum The maximum value of the seasonal day segmentation scheme corresponding to this maximum value is the final seasonal day segmentation scheme.

[0030] Step S3.8: The historical wind power output data is segmented according to the seasonal segmentation scheme finally determined in step S3.7 to form a dataset of Q seasons.

[0031] Step S4: Perform the following data statistics on the datasets for each season:

[0032] Divide the range 0 to 1 into K intervals, and count the number of wind power output values ​​at each same time in the data set for that season that fall into each interval, to obtain the wind power output distribution probability p for each time point in that season. r,t,q Where: r represents an interval, r = 1, ..., K; t is a certain time of day, t = 1, ..., m; q is the season number, q = 1, ..., Q;

[0033] The Markov transition matrix P(t) at each time step is statistically analyzed and calculated.

[0034]

[0035] In the formula: t represents a certain time of day; Let represent the transition probability between the two states at time t, where k represents the state at the previous time and l represents the state at the next time. When k = 1, the wind power output at the previous time is 0. When k takes the value [2, K+1], the wind power output at the previous time is ((k-2) / K, (k-1) / K] respectively. When l = 1, the wind power output at the next time is 0. When l takes the value [2, K+1], the wind power output at the next time is ((l-2) / K, (l-1) / K] respectively.

[0036] Step S5: Based on the results of steps S3 and S4, use the Metropolis-Hastington algorithm to generate simulation data for the whole year.

[0037] Furthermore, step S5 further includes:

[0038] Step S5.1: Determine which season of the year the date range for which data needs to be generated falls within, and select the wind power output distribution probability p for each moment of the corresponding season. r,t,q and the Markov transition matrix P(t);

[0039] Step S5.2: Simulate the data at each time point within the date range using the Metropolis-Hastington algorithm;

[0040] Step S5.3: Select different date ranges and repeat steps S5.1 to S5.2 to obtain the simulation data for the whole year.

[0041] Furthermore, during the simulation process in step S5.2, each piece of data obtained from the simulation is judged. If it satisfies the Markov transition matrix P(t), then the simulation meets the condition, the simulation data at that moment is recorded, and the simulation is performed at the next moment. If it does not satisfy the Markov transition matrix P(t), then the data is discarded and the simulation is regenerated and judged again until the condition is met.

[0042] Furthermore, the method includes step S6, which involves calculating a historical wind power output curve for each year based on historical wind power output data over many years; and calculating the Euler distance L between each pair of historical wind power output curves. w Then, by comparison, its maximum value L is obtained. w,max .

[0043] Furthermore, it also includes step S7, which verifies the simulation data for the entire year; this further includes:

[0044] Calculate the annual simulation data continuous output curve obtained in step S5; the shape and variation pattern of the annual simulation data continuous output curve should be similar to the historical wind power continuous output curve obtained in step S6.

[0045] Calculate the Eulerian distance L between the continuous power output curve of the annual simulation data and each historical continuous wind power output curve. v The largest value L v,max It should not exceed L w,max 1.5 times;

[0046] If all the above conditions are met, the simulation data for the whole year is valid simulation data; if not all the conditions are met, the simulation data for the whole year is discarded, and steps S5 to S7 are repeated.

[0047] Furthermore, it also includes step S8, repeating steps S5 to S7 to obtain valid annual simulation data for many years.

[0048] Preferably, the historical wind power output data collected in step S1 is at least three years' worth of data.

[0049] Preferably, in step S3.2, the annual wind power output is divided into three seasons (spring, autumn, summer, and winter) or four seasons (spring, summer, autumn, and winter) based on the principle of similarity.

[0050] Compared with the prior art, the beneficial technical effects of the present invention are as follows:

[0051] The present invention provides a method for generating wind power output simulation data based on historical data. This method statistically analyzes historical wind power output data over many years, dividing the output into different intervals according to output characteristics, and extracting the probability characteristics and Markov transition matrix for each interval. Based on the probabilities of each interval, multiple annual data sets are generated using the sequential Monte Carlo method, and corrected using historical continuous output curves. The resulting effective simulation data closely matches the probability distribution of local historical data and the local wind power output characteristics. It is suitable for computational studies such as Monte Carlo simulations requiring large amounts of data, resulting in better matching and greater objectivity of the research results. Attached Figure Description

[0052] Figure 1 This is an example of wind power continuous output curves for different regions and years. Detailed Implementation

[0053] This invention provides a method for generating wind power output simulation data based on historical data. The main purpose is to extract the characteristics of wind power output based on multi-year historical data, divide the year into different seasonal date segments according to the seasonal characteristics of wind power output, and generate simulation data of wind power output for the whole year based on the seasonal characteristics, for use in production simulation calculations and other simulation calculations.

[0054] Overall, the main steps and principles of the method for generating wind power output simulation data based on historical data according to the present invention are as follows:

[0055] Historical wind power output data is divided into several seasons using different date segmentation methods;

[0056] Calculate the expected wind power output for each season under each date segmentation method, select the date segmentation method with the largest deviation between expected values ​​for each season, and divide the historical wind power output data into several seasons using this date segmentation method.

[0057] The probability distribution of wind power output and the Markov transition matrix at each time point were obtained statistically.

[0058] Using the Metropolis-Hastings Algorithm, wind power output was simulated for each date and time in each season to obtain simulation data for the whole year.

[0059] Specifically, according to some embodiments, the method of generating wind power output simulation data based on historical data according to the present invention includes the following steps.

[0060] Step S1: Collect historical wind power output data for the local area over several years (at least three years) and normalize the data so that the data distribution is within the range of 0 to 1. The wind power output data is in tabular form, for example, which includes multiple wind power output values, and each wind power output value corresponds to a date and time.

[0061] Step S2 involves cleaning the data, removing erroneous data caused by sampling, recording, transmission, or abnormal weather, and replacing it with similar correct data. For example, if the wind power output values ​​of other adjacent locations are all high during a certain time period on a certain day, but are zero only at a certain moment, then this data can be determined to be erroneous. This data is then replaced with similar data based on data from adjacent moments or data from the same moment on an adjacent day, ensuring it largely conforms to natural laws.

[0062] Step S3 involves segmenting the year's dates according to the wind power output characteristics of different seasons. Analysis of historical wind power output data shows significant seasonal variations. Based on the principle of similarity, the year's wind power output can be divided into, for example, three seasons (spring / autumn, summer, and winter) or four seasons (spring, summer, autumn, and winter). Simulation mathematical models are then established for each date segment with similar characteristics, and simulations and data generation will follow accordingly. However, the seasonal boundaries should not be determined solely by fixed dates; they need to be obtained through statistical calculations of local historical wind power output data. This ensures that the simulation mathematical models and the final simulated data better match the local conditions.

[0063] Step S3 further includes:

[0064] Step S3.1: First, consider the local seasonal characteristics and set several seasonal transition periods. Specifically, for example, query the local annual temperature variations to analyze and determine the number of days and location of the seasonal transition periods within the year, dividing the year into several date segments as seasonal transition periods. These seasonal transition periods may include, for example, spring-summer, summer-autumn, autumn-winter, and winter-spring transition periods. The purpose of setting seasonal transition periods is to narrow down the range of seasonal division days. It can be assumed that the seasonal division days must fall within the date segments of the seasonal transition periods, thus facilitating subsequent enumeration of all possible permutations and combinations, separate calculations, and comparisons to obtain the optimal seasonal date segmentation scheme.

[0065] Step S3.2: Select a set of dates during the seasonal transition period to form a set of seasonal division date schemes, dividing the year into Q seasons. For example, the seasons are divided into spring, summer, autumn, and winter, with Q being 4; or when the wind power output in spring and autumn is similar, the seasons are divided into spring / autumn, summer, and winter, with Q being 3. The seasonal division date must fall within a certain season. The seasonal division date includes the end date of the previous season and the start date of the next season between any two adjacent seasons.

[0066] Step S3.3: Based on the seasonal segmentation scheme selected in step S3.2, integrate the historical wind power output data from multiple years that belong to the same season into a dataset for that season, forming a total of Q seasonal datasets, so that subsequent statistical calculations can be performed on the data of each season.

[0067] Step S3.4: Calculate the expected value E for each moment in the dataset for each season using the following formula. q,t and standard deviation σ q,t ;

[0068]

[0069]

[0070] In the formula: q is the season number (representing a certain season), q = 1, ..., Q; t is a certain time of day, t = 1, ..., m, m is, for example, 24; The wind power output value at time t on a certain day in season q; Let N be the average wind power output at time t for each day in the dataset of season q; N is the total number of days in the dataset of season q.

[0071] Step S3.5: For the expected value and standard deviation of the data for each season, calculate the Euclidean distance L between the seasons using the following formula. E,q Sum of standard deviation and Euler distance L σ,q :

[0072]

[0073]

[0074] In the formula: a and b represent two adjacent seasons respectively. The values ​​of a and b are: when q≠Q, a=q, b=q+1; when q=Q, a=q, b=1; m is the total number of moments in a day. For example, wind power output data is recorded once per hour, and there are 24 moments recorded per day, so m is 24.

[0075] The sum of the expected value Euler distance and the standard deviation Euler distance, L, is calculated using the following formula. sum :

[0076]

[0077] Step S3.6: During the seasonal transition period, select different seasonal day segmentation schemes, and repeat steps S3.2 to S3.5 to calculate the sum L of the expected value Eulerian distance and the standard deviation Eulerian distance corresponding to all possible seasonal day segmentation schemes. sum .

[0078] Step S3.7: By comparison, the sum L of the expected value Euler distance and the standard deviation Euler distance is obtained. sum The maximum value of the seasonal day segmentation scheme corresponding to this maximum value is the final determined optimal seasonal day segmentation scheme;

[0079] Step S3.8: The historical wind power output data is segmented according to the seasonal segmentation scheme finally determined in step S3.7 to form a dataset of Q seasons.

[0080] Step S4: Perform the following data statistics on the datasets for each season:

[0081] (1) Divide the range of 0 to 1 into K intervals (e.g., by a difference of 5% or 10%), and count the number of wind power output values ​​at each time point in the data set for the season that fall into each interval, to obtain the wind power output distribution probability p at each time point in the season. r,t,q Where: r represents an interval, r = 1, ..., K; t is a certain time of day, t = 1, ..., m; q is the season number, q = 1, ..., Q;

[0082] (2) The Markov transition matrix P(t) at each time step is obtained by statistical analysis and calculation.

[0083]

[0084] In the formula: t represents a certain time of day; Let represent the transition probability between the two states at time t, where k represents the state at the previous time and l represents the state at the next time. When k = 1, the wind power output at the previous time is 0. When k takes the value of [2, K+1], the wind power output at the previous time is ((k-2) / K, (k-1) / K] respectively. When l = 1, the wind power output at the next time is 0. When l takes the value of [2, K+1], the wind power output at the next time is ((l-2) / K, (l-1) / K] respectively. That is, the values ​​before and after are divided into 0, (0, 1 / K], (1 / K, 2 / K], ..., ((K-1) / K, 1] respectively, and the above matrix is ​​obtained by arranging and combining them in sequence.

[0085] Step S5: Based on the results of steps S3 and S4, use the Metropolis-Hastington algorithm to generate simulation data for the whole year.

[0086] More specifically, step S5 further includes:

[0087] Step S5.1: Traverse the entire year. First, determine which season of the year the date range for which data needs to be generated falls within, and select the wind power output distribution probability p for each moment in the corresponding season. r,t,q And the Markov transition matrix P(t).

[0088] Step S5.2: Based on the data statistics results from steps S3 and S4, the Metropolis-Hastings algorithm is used to simulate the data at each time point within the date range. During the simulation in step S5.2, each data point obtained from the simulation is evaluated. If it satisfies the Markov transition matrix P(t), the simulation meets the condition, the simulation data V(t) at that time is recorded, and the simulation proceeds to the next time point. If it does not satisfy the Markov transition matrix P(t), the data is discarded, and the simulation is regenerated and evaluated again until the condition is met.

[0089] Step S5.3: Select different date ranges and repeat steps S5.1 to S5.2 to finally obtain the simulation data V for the whole year. year .

[0090] Furthermore, steps S6 and S7 are included to verify the simulation data.

[0091] Step S6: Based on the historical wind power output data (processed in steps S1 and S2) over many years, calculate a historical continuous wind power output curve C for each year's wind power output data. w,s ; Calculate the Euler distance L between each of the two historical wind power continuous output curves. w Then, by comparison, its maximum value L is obtained. w,max .

[0092] Please see Figure 1 Any curve in the graph, the continuous wind power output curve is plotted by arranging the wind power output values ​​of a year from largest to smallest. The vertical axis of the graph is the wind power output value, and the horizontal axis is the sequence number of the wind power output values ​​after sorting (each wind power output value corresponds to a time, so in this graph there are approximately 365*24=8760 wind power output values).

[0093] The Euler distance L between the two historical continuous wind power output curves w The calculation is performed using the following formula:

[0094]

[0095] In the formula: n is the total number of serial numbers, for example, 8760; α and β represent the two curves respectively; This represents the difference in wind power output at the same horizontal coordinate position.

[0096] Step S7, verifying the simulation data for the whole year; it further includes:

[0097] (1) Calculate the annual simulation data V obtained in step S5. year The annual simulation data continuous output curve C year,s ;Continuous output curve C of simulation data throughout the year year,s The shape and variation pattern should correspond to the historical continuous wind power output curve C obtained in step S6. w,s resemblance.

[0098] Please see Figure 1 The three solid lines represent the three-year continuous wind power output curves of location A, while the two dashed lines represent the two-year continuous wind power output curves of location B. It can be seen that the shape and variation patterns of the continuous wind power output curves of the same location in different years (more specifically, the slope of each position of the curve, the change of the slope, the position of the intersection of the curve with the horizontal and vertical axes, the integral area of ​​the curve, etc.) are roughly similar, while there are more obvious differences in different regions. Therefore, it is possible to judge whether the simulation data is successful and whether it matches the local actual situation.

[0099] (2) Calculate the Euler distance L between the continuous power output curve of the annual simulation data and each historical wind power continuous power output curve. v (The formula for calculating the Euler distance is similar to that above and will not be repeated here), where the maximum value L v,max It should not exceed L w,max 1.5 times.

[0100] Only when all conditions (1) and (2) above are met can the annual simulation data V be considered valid. year For this to be considered valid simulation data; if not all conditions are met, then the annual simulation data V is considered valid. year This is invalid data; discard the entire year's simulation data V.year Repeat steps S5 to S7 to re-perform simulation generation and verification (the data generated in each simulation will have deviations, so there will be cases where the verification passes or fails).

[0101] Furthermore, it also includes step S8, repeating steps S5 to S7 to obtain valid annual simulation data for many years, thereby meeting the needs of research calculations such as Monte Carlo simulation calculations.

[0102] In summary, the method for generating wind power output simulation data based on historical data of the present invention divides wind power output into different intervals according to output characteristics by statistically analyzing historical wind power output data over many years, and extracts the probability characteristics and Markov state transition matrix of each interval. Based on the probability of each interval, multiple annual data sets are generated using the sequential Monte Carlo method, and corrected using historical continuous output curves. The effective simulation data that meets the requirements can better conform to the probability distribution of local historical data and the output characteristics of local wind power. It is suitable for computational studies such as Monte Carlo simulation calculations that require a large amount of data, and the research results have better matching and greater objectivity.

Claims

1. A method for generating simulated wind power output data based on historical data, characterized in that, Includes the following steps: Historical wind power output data is divided into several seasons; The probability distribution of wind power output and the Markov transition matrix at each time point were obtained statistically. Using the Metropolis-Hastington algorithm, the wind power output of each date and time in each season was simulated to obtain simulation data for the whole year. Among them, the historical wind power output data is divided into several seasons through the following step S3; Step S3: Divide the year into segments according to the wind power output characteristics of different seasons; Step S3 further includes: Step S3.1: First, consider the local seasonal characteristics and set the seasonal transition period; Step S3.2: Select a set of dates during the seasonal transition period to form a set of seasonal day division schemes, dividing the year into Q seasons; Step S3.3: Integrate the historical wind power output data from multiple years that belong to the same season into a dataset for that season; Step S3.4: Calculate the expected value E for each moment in the dataset for each season using the following formula. q,t and standard deviation σ q,t ; In the formula: q is the season number, q=1,…,Q; t is a certain time of day; The wind power output value at time t on a certain day in season q; Let N be the average wind power output at time t for each day in the dataset of season q; N is the total number of days in the dataset of season q. Step S3.5: Calculate the expected Euler distance L between seasons using the following formula. E,q Sum of standard deviation and Euler distance L σ,q : In the formula: a and b represent two adjacent seasons respectively. When q≠Q, a=q, b=q+1; when q=Q, a=q, b=1; m is the total number of moments in a day. The sum of the expected value Euler distance and the standard deviation Euler distance, L, is calculated using the following formula. sum : ； Step S3.6: During the seasonal transition period, select different seasonal day segmentation schemes, and repeat steps S3.2 to S3.5 to calculate the sum L of the expected value Eulerian distance and the standard deviation Eulerian distance corresponding to all possible seasonal day segmentation schemes. sum ; Step S3.7: By comparison, the sum L of the expected value Euler distance and the standard deviation Euler distance is obtained. sum The maximum value of the seasonal day segmentation scheme corresponding to this maximum value is the final seasonal day segmentation scheme. Step S3.8: The historical wind power output data is segmented according to the seasonal segmentation scheme finally determined in step S3.7 to form a dataset of Q seasons.

2. The method for generating simulated wind power output data based on historical data according to claim 1, characterized in that, Steps S1 and S2 are included before step S3; Step S1: Collect historical wind power output data for the local area over many years and normalize the data; Step S2: Clean the data, remove erroneous data, and replace it with correct data; Step S3 is followed by steps S4 and S5; Step S4: Perform the following data statistics on the datasets for each season: Divide the range 0 to 1 into K intervals, and count the number of wind power output values ​​at each same time in the data set for that season that fall into each interval, to obtain the wind power output distribution probability p for each time of that season. r,t,q Where: r represents an interval, r=1,…,K; t is a certain time of day, t=1,…,m; q is the season number, q=1,…,Q; The Markov transition matrix P(t) at each time step is statistically analyzed and calculated. In the formula: t represents a certain time of day; Let represent the transition probability between the two states at time t. k Indicates the state at the previous moment. l Indicates the state at the next moment; k =1 indicates that the wind power output was 0 at the previous moment. k When the value is in [2, K+1], it means that the wind power output value at the previous moment was (( k -2) / K,( k -1) / K]; l =1 indicates that the wind power output is 0 at the next moment. l When the value is in [2, K+1], it means that the wind power output value at the next moment is (( l -2) / K,( l -1) / K]; Step S5: Based on the results of steps S3 and S4, use the Metropolis-Hastington algorithm to generate simulation data for the whole year.

3. The method for generating simulated wind power output data based on historical data according to claim 2, characterized in that, Step S5 further includes: Step S5.1: Determine which season of the year the date range for which data needs to be generated falls within, and select the wind power output distribution probability p for each moment of the corresponding season. r,t,q and the Markov transition matrix P(t); Step S5.2: Simulate the data at each time point within the date range using the Metropolis-Hastington algorithm; Step S5.3: Select different date ranges and repeat steps S5.1 to S5.2 to obtain the simulation data for the whole year.

4. The method for generating wind power output simulation data based on historical data according to claim 3, characterized in that, During the simulation in step S5.2, each data obtained from the simulation is judged. If it satisfies the Markov transition matrix P(t), then the simulation satisfies the condition. The simulation data at that moment is recorded, and the simulation is performed at the next moment. If the Markov transition matrix P(t) is not satisfied, the data is discarded and the simulation is repeated until the condition is met.

5. The method for generating simulated wind power output data based on historical data according to any one of claims 2-4, characterized in that, The process also includes step S6, which involves calculating a historical wind power output curve for each year based on historical wind power output data over many years; and calculating the Euler distance L between each pair of historical wind power output curves. w Then, by comparison, its maximum value L is obtained. w,max .

6. The method for generating simulated wind power output data based on historical data according to claim 5, characterized in that, It also includes step S7, which verifies the simulation data for the whole year; It further includes: Calculate the annual simulation data continuous output curve obtained in step S5; the shape and variation pattern of the annual simulation data continuous output curve should be similar to the historical wind power continuous output curve obtained in step S6. Calculate the Eulerian distance L between the continuous power output curve of the annual simulation data and each historical continuous wind power output curve. v The largest value L v,max It should not exceed L w,max 1.5 times; If all the above conditions are met, the simulation data for the whole year is valid simulation data; if not all the conditions are met, the simulation data for the whole year is discarded, and steps S5 to S7 are repeated.

7. The method for generating simulated wind power output data based on historical data according to claim 6, characterized in that, It also includes step S8, which repeats steps S5 to S7 to obtain years of valid annual simulation data.

8. The method for generating simulated wind power output data based on historical data according to any one of claims 2-4, characterized in that, The historical wind power output data collected in step S1 must be at least three years' worth of data.

9. The method for generating simulated wind power output data based on historical data according to any one of claims 1-4, characterized in that, In step S3.2, the annual wind power output is divided into three seasons (spring, autumn, summer, and winter) or four seasons (spring, summer, autumn, and winter) based on the principle of similarity.

Images