A method and system for generating wind power prediction scenarios

Abstract

The present invention provides a method and system for generating a wind power prediction scenario, comprising: obtaining a current prediction power sequence, a historical prediction power sequence, and a historical measured power sequence; based on the current prediction power sequence, the historical prediction power sequence, and the historical measured power sequence, sequentially calculating the current prediction power at two given adjacent moments and the power prediction error at the previous moment in the two adjacent moments in each scenario, to obtain a power prediction error sequence for each scenario; based on the current prediction power sequence and the power prediction error sequence for each scenario, generating a wind power prediction power sequence for each scenario. The present invention takes into account the time correlation of the power prediction errors at adjacent moments, so that the generated wind power prediction scenario well retains the temporal nature of the wind power prediction, thereby improving the reliability of the wind power prediction scenario.

Description

Technical Field

[0001] The invention belongs to the field of power system optimization and scheduling, and specifically relates to a method and system for generating wind power prediction scenarios. Background Art

[0002] Due to the strong randomness and uncertainty of wind power, existing deterministic scheduling methods based on point forecast results are no longer sufficient to meet the needs of optimizing grid day-ahead scheduling. A stochastic optimization scheduling method based on forecast scenarios is one feasible approach to improving the rationality of grid day-ahead generation planning decisions. This method relies on a series of wind power forecast scenarios and, through unit commitment optimization, derives the optimal generation plan under all forecast scenarios. This plan optimizes power generation economics while ensuring safe grid operation. Because stochastic optimization scheduling methods account for the randomness of wind power, scheduling decisions are more scientific and reasonable. Wind power forecast scenarios are a key input for stochastic optimization scheduling decisions. Currently, grid dispatching departments generally use a method to generate forecast scenarios based on point forecast results. This method relies on independent sampling at each time interval based on the probability distribution of historical forecast errors. However, because it fails to consider the temporal correlation between forecast errors, the temporal characteristics of the generated power forecast scenarios are difficult to guarantee, impacting their credibility and leading to irrational scheduling decisions. Therefore, addressing these issues in the existing technology is a challenge for those skilled in the art. Summary of the Invention

[0003] To overcome the above-mentioned deficiencies of the prior art, the present invention provides a method for generating a wind power prediction scenario, comprising:

[0004] Obtain the current predicted power series, historical predicted power series, and historical measured power series;

[0005] Based on the current predicted power sequence, the historical predicted power sequence, and the historical measured power sequence, sequentially calculating the power prediction error at the next moment in each scenario under the given current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments, to obtain a power prediction error sequence for each scenario;

[0006] Based on the current predicted power sequence and the power prediction error sequence in each scenario, a wind power predicted power sequence in each scenario is generated.

[0007] Preferably, the calculation of the power prediction error sequence includes:

[0008] Based on the historical predicted power sequence and the historical measured power sequence, constructing a joint distribution variable of predicted power and power prediction error at adjacent historical moments;

[0009] Cluster the variables of the joint distribution of the predicted power and power prediction error at adjacent historical moments, and obtain the cluster centers of each category and the probability density function of the joint distribution of the predicted power and power prediction error at adjacent historical moments;

[0010] Based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, the power prediction error at the next moment is calculated in sequence under the conditions of the current predicted power of two given adjacent moments and the power prediction error of the previous moment in the two adjacent moments.

[0011] Preferably, based on the current predicted power sequence, the cluster centers of each category, and the predicted power and power prediction error joint distribution probability density function of each category at historical adjacent moments, sequentially calculating the power prediction error at the next moment given the current predicted power of two adjacent moments and the power prediction error at the previous moment in the two adjacent moments, including:

[0012] Calculate the power prediction errors at the first and second moments in the power prediction error sequence based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction errors at historical adjacent moments of each category;

[0013] Based on the current predicted power sequence, the cluster centers of each category, the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, and the power prediction error at the first moment and the second moment in the power prediction error sequence, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence is calculated in sequence given the current predicted power at two adjacent moments and the power prediction error at the previous moment between the two adjacent moments;

[0014] The conditional probability distribution of the power prediction error at the third moment and each subsequent moment in the power prediction error sequence is randomly sampled in sequence to obtain the power prediction error at the third moment and each subsequent moment.

[0015] Preferably, based on the current predicted power sequence, the cluster centers of each category, and the predicted power and power prediction error joint distribution probability density function of each category at adjacent historical moments, calculating the power prediction error at the first moment and the second moment in the power prediction error sequence includes:

[0016] Based on the current predicted power sequence, constructing a high-dimensional vector of the first adjacent moment; the high-dimensional vector of the first adjacent moment is composed of the current predicted power at the first moment, the current predicted power at the second moment, the initial value of the power prediction error set at the first moment, and the initial value of the power prediction error set at the second moment;

[0017] Based on the various cluster centers, determining the category to which the high-dimensional vector at the first adjacent moment belongs;

[0018] Randomly sample the predicted power and power prediction error joint distribution probability density function of the historical adjacent moments of the category to which the high-dimensional vector of the first adjacent moment belongs to obtain the power prediction error at the first moment and the second moment in the power prediction error sequence.

[0019] Preferably, based on the current predicted power sequence, the cluster centers of each category, the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, and the power prediction error at the first moment and the second moment in the power prediction error sequence, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence is calculated in sequence under the conditions of the current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments, including:

[0020] Based on the current predicted power sequence and the power prediction error at the second moment in the power prediction error sequence, sequentially generating high-dimensional vectors for each adjacent moment after the first adjacent moment; the high-dimensional vectors for each adjacent moment after the first adjacent moment are composed of the current predicted power of the two adjacent moments, the power prediction error at the previous moment between the two adjacent moments, and the power prediction error at the next moment between the two adjacent moments set to an initial value;

[0021] Based on the cluster centers of the categories, sequentially determining the categories to which the high-dimensional vectors at each adjacent time after the first adjacent time belong;

[0022] Based on the joint distribution probability density function of the predicted power and power prediction error of the historical adjacent moments of the category to which the high-dimensional vectors at each adjacent moment after the first adjacent moment belong, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence is calculated in sequence.

[0023] Preferably, generating a wind power prediction power sequence for each scenario based on the current prediction power sequence and the power prediction error sequence for each scenario includes:

[0024] The power prediction error under each scenario is added to the current predicted power at the corresponding moment to obtain the wind power prediction power sequence under each scenario.

[0025] Based on the same concept, the present invention provides a wind power prediction scenario generation system, comprising:

[0026] Data acquisition module, used to obtain the current predicted power series, historical predicted power series and historical measured power series;

[0027] A power prediction error generation module is used to calculate, in each scenario, the power prediction error at the next moment given the current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments based on the current predicted power sequence, the historical predicted power sequence, and the historical measured power sequence, to obtain a power prediction error sequence for each scenario;

[0028] The power prediction scenario generation module is used to generate a wind power prediction power sequence for each scenario based on the current prediction power sequence and the power prediction error sequence for each scenario.

[0029] Preferably, the power prediction error generating module includes:

[0030] A joint distribution variable construction module is used to construct a joint distribution variable of the predicted power and power prediction error at adjacent historical moments based on the historical predicted power sequence and the historical measured power sequence;

[0031] A clustering module is used to cluster the variables of the joint distribution of the predicted power and power prediction error at adjacent historical moments, and obtain the cluster centers of each category and the probability density function of the joint distribution of the predicted power and power prediction error at adjacent historical moments;

[0032] A calculation module is used to calculate the power prediction error at the next moment under the conditions of the current predicted power at two given adjacent moments and the power prediction error at the previous moment in the two adjacent moments based on the current predicted power sequence, the cluster centers of each category and the joint distribution probability density function of the predicted power and power prediction error at the historical adjacent moments of each category.

[0033] Preferably, the computing module includes:

[0034] A first calculation submodule is used to calculate the power prediction error at the first moment and the second moment in the power prediction error sequence based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction error at historical adjacent moments of each category;

[0035] A second calculation submodule is used to calculate, based on the current predicted power sequence, the cluster centers of each category, the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, and the power prediction error at the first moment and the second moment in the power prediction error sequence, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence under the conditions of the current predicted power at two adjacent moments and the power prediction error at the previous moment between the two adjacent moments;

[0036] The third calculation submodule is used to randomly sample the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence in sequence to obtain the power prediction error at the third moment and subsequent moments.

[0037] Preferably, the first calculation submodule includes:

[0038] an initial high-dimensional vector construction unit, configured to construct a high-dimensional vector at a first adjacent moment based on the current predicted power sequence; the high-dimensional vector at the first adjacent moment being composed of the current predicted power at the first moment, the current predicted power at the second moment, an initial value set for the power prediction error at the first moment, and an initial value set for the power prediction error at the second moment;

[0039] An initial high-dimensional vector classification unit, configured to determine the category to which the high-dimensional vector at the first adjacent moment belongs based on the cluster centers of each type;

[0040] The initial adjacent moment power prediction error calculation unit is used to randomly sample the predicted power and power prediction error joint distribution probability density function of the historical adjacent moments of the category to which the high-dimensional vector of the first adjacent moment belongs, and obtain the power prediction error of the first moment and the second moment in the power prediction error sequence.

[0041] Preferably, the second calculation submodule includes:

[0042] Other high-dimensional vector construction units are used to sequentially generate high-dimensional vectors for each adjacent moment after the first adjacent moment based on the current predicted power sequence and the power prediction error at the second moment in the power prediction error sequence; the high-dimensional vectors for each adjacent moment after the first adjacent moment are composed of the current predicted power of the two adjacent moments, the power prediction error at the previous moment between the two adjacent moments, and the power prediction error at the next moment between the two adjacent moments. Initial values ​​are set;

[0043] Other high-dimensional vector classification units are used to determine, based on the cluster centers of the categories, the categories to which the high-dimensional vectors at each adjacent moment after the first adjacent moment belong;

[0044] The power prediction error calculation unit at other moments is used to calculate the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence based on the joint distribution probability density function of the predicted power and power prediction error of the historical adjacent moments of the category to which the high-dimensional vectors of each adjacent moment after the first adjacent moment belong.

[0045] Compared with the closest prior art, the present invention has the following beneficial effects:

[0046] The present invention provides a method and system for generating a wind power prediction scenario, comprising: obtaining a current prediction power sequence, a historical prediction power sequence, and a historical measured power sequence; based on the current prediction power sequence, the historical prediction power sequence, and the historical measured power sequence, sequentially calculating the current prediction power at two given adjacent moments and the power prediction error at the previous moment in the two adjacent moments in each scenario, to obtain a power prediction error sequence for each scenario; based on the current prediction power sequence and the power prediction error sequence for each scenario, generating a wind power prediction power sequence for each scenario. The present invention takes into account the time correlation of the power prediction errors at adjacent moments, so that the generated wind power prediction scenario well retains the temporal nature of the wind power prediction, thereby improving the reliability of the wind power prediction scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a method for generating a wind power prediction scenario provided by the present invention;

[0048] Figure 2 A schematic diagram of a wind power prediction scenario generation system provided by the present invention;

[0049] Figure 3 This is a scatter plot of wind power prediction samples at time t and time t+1 in the eight cluster sample sets provided in the embodiment of the present invention;

[0050] Figure 4 This is a scatter plot of wind power prediction error samples at time t and time t+1 in the eight cluster sample sets provided in the embodiment of the present invention;

[0051] Figure 5 A comparison diagram of wind power prediction scenario results when time correlation is considered and when time correlation is not considered provided in an embodiment of the present invention;

[0052] Figure 6 A comparison diagram of autocorrelation coefficient results of wind power prediction scenarios with and without considering time correlation is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0054] Example 1:

[0055] The embodiment of the present invention discloses a method for generating a wind power prediction scenario. Figure 1 As shown, including:

[0056] S1 obtains the current predicted power sequence, the historical predicted power sequence and the historical measured power sequence;

[0057] S2, based on the current predicted power sequence, the historical predicted power sequence, and the historical measured power sequence, sequentially calculates the power prediction error at the next moment in each scenario under the given current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments, to obtain a power prediction error sequence for each scenario;

[0058] S3 generates a wind power prediction power sequence for each scenario based on the current prediction power sequence and the power prediction error sequence for each scenario.

[0059] Among them, S1 obtains the current predicted power series, historical predicted power series, and historical measured power series, including:

[0060] Obtain historical data on wind power, including power point prediction sequences Measured power sequence and power prediction error series Where T0 is the total number of moments of historical data, and the prediction error at time t is

[0061] S2, based on the current predicted power sequence, the historical predicted power sequence, and the historical measured power sequence, sequentially calculates the power prediction error at the next moment in each scenario under the given current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments, to obtain the power prediction error sequence for each scenario, including:

[0062] S2-1 Clustering of Wind Power Forecast and Forecast Error Samples

[0063] Combine the prediction power and prediction error at time t and time t+1 into a joint distribution variable Θ t ,Right now Assume that the sample set composed of the joint distribution variables at all times is Θ0, that is:

[0064]

[0065] Given the number of cluster categories K, the k-means clustering method is used to cluster the sample set Θ0. Assume that the sample sets of the joint distribution variables under each category are {Θ1, Θ2, ..., Θ K}, the cluster centers under each category are {C1,C2,...,C K}.

[0066] S2-2 Estimation of the joint distribution probability density function of prediction power and prediction error

[0067] For each category, the sample set {Θ1,Θ2,...,Θ K}, the probability density function is estimated by using the mixed Gaussian model to obtain the probability density function of the joint distribution of prediction power and prediction error {b1(x),b2(x),...,b K (x)}, where the probability density function corresponding to the kth joint distribution variable sample set is:

[0068]

[0069] Among them, M is the number of Gaussian distributions, c m is the weight of the mth Gaussian distribution, and the sum of all M weights is 1, that is, f(x;μ m ,Σ m ) is the probability density function of the mth Gaussian distribution, which is as follows:

[0070]

[0071] The mature Expectation-Maximization algorithm (EM) is used to estimate the parameters of the Gaussian mixture model corresponding to each category. m ,μ m ,Σ m}, and obtain the probability density function of the joint distribution of prediction power and prediction error under each category {b1(x),b2(x),...,b K (x)}.

[0072] S2-3 Wind power forecast scenario generation

[0073] Assume that you need to generate the future from t1 to t T N wind power forecast scenarios within the time period, and the future time from t1 to t T Wind power point prediction series within the time period and the probability density function of the joint distribution of prediction power and prediction error obtained based on historical data {b1(x),b2(x),...,b K (x)}, the scene generation steps are as follows:

[0074] S2-3-1 Assume that the nth random scenario is currently being generated. At the initial moment t1, the initial values ​​are set based on the point predicted power and power prediction error at moments t1 and t2 to form a high-dimensional vector The power prediction error at time t1 and t2 is set to 0 initially. With the K-category cluster centers {C1,C2,...,C K}, if the Euclidean distance with the center of the k1 class is the smallest, then it is considered Belongs to the k1th class. The probability density function of the prediction power and prediction error based on the k1th class Perform random sampling to obtain prediction error samples at time t1 and t2

[0075] S2-3-2 Assume that the current time is t l ,2≤l≤T-1, based on t l and t l+1 The predicted power at time t l The prediction error sample at time and t l+1 The initial value of the moment prediction error is set to form a high-dimensional vector t l+1 The initial value of the moment prediction error is set to 0, and the calculation With the K-category cluster centers {C1,C2,...,C K}, if the Euclidean distance with the kth l The center distance of the cluster is the smallest, then it is considered Belongs to the kth l Class. Based on the kth l Probability density functions of prediction power and prediction error for the class Calculate in known In case t l+1 The conditional probability distribution function a obeyed by the moment prediction error l+1 (z|y), a l+1 (z|y) can be estimated using a Gaussian mixture model:

[0076]

[0077] in, is the probability density function of the mth Gaussian model. The parameters of the mixed Gaussian model are calculated as follows:

[0078]

[0079]

[0080]

[0081] in, Corresponding to μ in step 2 m ,Σ m Subvectors and block submatrices in :

[0082]

[0083]

[0084] Based on the obtained conditional probability distribution function a l+1 (z|y), obtained by random sampling l+1 The prediction error sample at time

[0085] S2-3-3 repeats step S2-3-2 until t is obtained T The prediction error sample at time So we can get the future t1 to t T The nth wind power forecast error scenario within the time period

[0086] S2-3-4 determines whether n is equal to N. If not, repeat steps S2-3-1 to S2-3-3. If yes, get the future time from t1 to t T N wind power forecast error scenarios within a time period:

[0087]

[0088] S2-3-5 Based on the wind power point prediction results and prediction error scenarios, the future time from t1 to t T A set of N wind power prediction scenarios within a period

[0089]

[0090] Example 2:

[0091] This embodiment of the present invention discloses a method tested using 15-minute data from a wind farm over six months. First, samples of wind power forecasts and forecast errors at adjacent moments were clustered into eight clusters. A Gaussian mixture model containing three Gaussian distributions was used to estimate the probability density function of each sample. Figure 3 and Figure 4 These are the sample scatter plots of wind power prediction and prediction error at time t and time t+1 in 8 cluster sample sets, Figure 4 The thin solid line in is the joint probability distribution curve fitted by the mixed Gaussian model. Figure 3 and Figure 4 It can be seen that the clustering algorithm divides the wind power prediction power and prediction error at adjacent moments very well, such as Figure 3 In the first and last subgraphs of , although the prediction powers of samples in categories 1 and 8 overlap, Figure 4 From the first and last sub-graphs, we can see that the prediction errors of the two categories are significantly different.

[0092] Figure 5The wind power prediction scenario results when considering time correlation and not considering time correlation at 96 moments on a certain day were compared. The number of predicted power scenarios generated was 20. In the figure, the gray dotted line represents the measured power, the black thick solid line represents the point predicted power, and the gray thin solid line represents the 20 power prediction scenarios. It can be found that when time correlation is considered, the prediction scenario has obvious temporal sequence, that is, the predicted power change trends at adjacent moments are relatively consistent, and there are fewer cases of sudden changes in predicted power; when time correlation is not considered, the prediction scenario has poor temporal sequence, that is, the predicted power change trends at adjacent moments are often opposite, the predicted power shows a "sawtooth" fluctuation, and the probability of sudden changes is high. Figure 6 The autocorrelation coefficients of wind power forecast scenarios at 96 moments on a certain day are compared when time correlation is considered and when time correlation is not considered. The gray dotted line represents the autocorrelation coefficient of the point prediction power, and the gray solid line represents the autocorrelation coefficient of the 20 power forecast scenarios. The results show that the autocorrelation of the power forecast scenarios generated when time correlation is considered at different time delays is closer to the autocorrelation of the point prediction power. Figure 5 and Figure 6 The results in this paper prove that the proposed method can fully take into account the time series characteristics of wind power, and the generated power forecast scenario is more reasonable. The results verify the effectiveness of the method.

[0093] Example 3:

[0094] The embodiment of the present invention discloses a wind power prediction scenario generation system. Figure 2 As shown, including:

[0095] Data acquisition module, used to obtain the current predicted power series, historical predicted power series and historical measured power series;

[0096] A power prediction error generation module is used to calculate, in each scenario, the power prediction error at the next moment given the current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments based on the current predicted power sequence, the historical predicted power sequence, and the historical measured power sequence, to obtain a power prediction error sequence for each scenario;

[0097] The power prediction scenario generation module is used to generate a wind power prediction power sequence for each scenario based on the current prediction power sequence and the power prediction error sequence for each scenario.

[0098] The power prediction error generation module includes:

[0099] A joint distribution variable construction module is used to construct a joint distribution variable of the predicted power and power prediction error at adjacent historical moments based on the historical predicted power sequence and the historical measured power sequence;

[0100] A clustering module is used to cluster the variables of the joint distribution of the predicted power and power prediction error at adjacent historical moments, and obtain the cluster centers of each category and the probability density function of the joint distribution of the predicted power and power prediction error at adjacent historical moments;

[0101] A calculation module is used to calculate the power prediction error at the next moment under the conditions of the current predicted power at two given adjacent moments and the power prediction error at the previous moment in the two adjacent moments based on the current predicted power sequence, the cluster centers of each category and the joint distribution probability density function of the predicted power and power prediction error at the historical adjacent moments of each category.

[0102] Computing module, including:

[0103] A first calculation submodule is used to calculate the power prediction error at the first moment and the second moment in the power prediction error sequence based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction error at historical adjacent moments of each category;

[0104] A second calculation submodule is used to calculate, based on the current predicted power sequence, the cluster centers of each category, the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, and the power prediction error at the first moment and the second moment in the power prediction error sequence, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence under the conditions of the current predicted power at two adjacent moments and the power prediction error at the previous moment between the two adjacent moments;

[0105] The third calculation submodule is used to randomly sample the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence in sequence to obtain the power prediction error at the third moment and subsequent moments.

[0106] The first computing submodule includes:

[0107] an initial high-dimensional vector construction unit, configured to construct a high-dimensional vector at a first adjacent moment based on the current predicted power sequence; the high-dimensional vector at the first adjacent moment being composed of the current predicted power at the first moment, the current predicted power at the second moment, an initial value set for the power prediction error at the first moment, and an initial value set for the power prediction error at the second moment;

[0108] An initial high-dimensional vector classification unit, configured to determine the category to which the high-dimensional vector at the first adjacent moment belongs based on the cluster centers of each type;

[0109] The initial adjacent moment power prediction error calculation unit is used to randomly sample the predicted power and power prediction error joint distribution probability density function of the historical adjacent moments of the category to which the high-dimensional vector of the first adjacent moment belongs, and obtain the power prediction error of the first moment and the second moment in the power prediction error sequence.

[0110] The second computing submodule includes:

[0111] Other high-dimensional vector construction units are used to sequentially generate high-dimensional vectors for each adjacent moment after the first adjacent moment based on the current predicted power sequence and the power prediction error at the second moment in the power prediction error sequence; the high-dimensional vectors for each adjacent moment after the first adjacent moment are composed of the current predicted power of the two adjacent moments, the power prediction error at the previous moment between the two adjacent moments, and the power prediction error at the next moment between the two adjacent moments. Initial values ​​are set;

[0112] Other high-dimensional vector classification units are used to determine, based on the cluster centers of the categories, the categories to which the high-dimensional vectors at each adjacent moment after the first adjacent moment belong;

[0113] The power prediction error calculation unit at other moments is used to calculate the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence based on the joint distribution probability density function of the predicted power and power prediction error of the historical adjacent moments of the category to which the high-dimensional vectors of each adjacent moment after the first adjacent moment belong.

[0114] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0115] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0116] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit its scope of protection. Although the present application has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that after reading this application, those skilled in the art may still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the application.

Claims

1. A method for generating a wind power prediction scenario, characterized in that: include: Obtain the current predicted power series, historical predicted power series, and historical measured power series; Based on the current predicted power sequence, the historical predicted power sequence, and the historical measured power sequence, sequentially calculating the power prediction error at the next moment in each scenario under the given current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments, to obtain a power prediction error sequence for each scenario; Generate a wind power prediction power sequence for each scenario based on the current prediction power sequence and the power prediction error sequence for each scenario; The calculation of the power prediction error sequence in each scenario includes: Based on the historical predicted power sequence and the historical measured power sequence, constructing a joint distribution variable of predicted power and power prediction error at adjacent historical moments; Cluster the joint distribution variables of the predicted power and power prediction error at adjacent historical moments, and obtain the probability density function of the joint distribution of the predicted power and power prediction error at adjacent historical moments in each category; Based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, the power prediction error at the next moment is calculated in sequence under the conditions of the current predicted power of two given adjacent moments and the power prediction error of the previous moment in the two adjacent moments.

2. The method according to claim 1, wherein The method of calculating the power prediction error at a later moment based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction error at historical adjacent moments of each category, under the condition that the current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments are given, includes: Calculate the power prediction errors at the first and second moments in the power prediction error sequence based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction errors at historical adjacent moments of each category; Based on the current predicted power sequence, the cluster centers of each category, the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, and the power prediction error at the first moment and the second moment in the power prediction error sequence, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence is calculated in sequence given the current predicted power at two adjacent moments and the power prediction error at the previous moment between the two adjacent moments; The conditional probability distribution of the power prediction error at the third moment and each subsequent moment in the power prediction error sequence is randomly sampled in sequence to obtain the power prediction error at the third moment and each subsequent moment.

3. The method according to claim 2, wherein The calculating the power prediction errors at the first moment and the second moment in the power prediction error sequence based on the current predicted power sequence, the cluster centers of each category, and the predicted power and power prediction error joint distribution probability density function of each category at historical adjacent moments includes: Based on the current predicted power sequence, constructing a high-dimensional vector of the first adjacent moment; the high-dimensional vector of the first adjacent moment is composed of the current predicted power at the first moment, the current predicted power at the second moment, the initial value of the power prediction error set at the first moment, and the initial value of the power prediction error set at the second moment; Based on the cluster centers of each category, determining the category to which the high-dimensional vector at the first adjacent moment belongs; Randomly sample the predicted power and power prediction error joint distribution probability density function of the historical adjacent moments of the category to which the high-dimensional vector of the first adjacent moment belongs to obtain the power prediction error at the first moment and the second moment in the power prediction error sequence.

4. The method according to claim 2, wherein The method is based on the current predicted power sequence, the cluster centers of each category, the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, and the power prediction error at the first moment and the second moment in the power prediction error sequence, and sequentially calculating the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence under the conditions of the current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments, including: Based on the current predicted power sequence and the power prediction error at the second moment in the power prediction error sequence, sequentially generating high-dimensional vectors for each adjacent moment after the first adjacent moment; the high-dimensional vectors for each adjacent moment after the first adjacent moment are composed of the current predicted power of the two adjacent moments, the power prediction error at the previous moment between the two adjacent moments, and the power prediction error at the next moment between the two adjacent moments set to an initial value; Based on the cluster centers of the categories, sequentially determining the categories to which the high-dimensional vectors at each adjacent time after the first adjacent time belong; Based on the joint distribution probability density function of the predicted power and power prediction error of the historical adjacent moments of the category to which the high-dimensional vectors at each adjacent moment after the first adjacent moment belong, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence is calculated in sequence.

5. The method according to claim 1, wherein The generating of the wind power prediction power sequence for each scenario based on the current prediction power sequence and the power prediction error sequence for each scenario includes: The power prediction error under each scenario is added to the current predicted power at the corresponding moment to obtain the wind power prediction power sequence under each scenario.

6. A wind power prediction scenario generation system, characterized in that: include: Data acquisition module, used to obtain the current predicted power series, historical predicted power series and historical measured power series; A power prediction error generation module is used to calculate, in each scenario, the power prediction error at the next moment given the current predicted power at two adjacent moments and the power prediction error at the previous moment in the two adjacent moments based on the current predicted power sequence, the historical predicted power sequence, and the historical measured power sequence, to obtain a power prediction error sequence for each scenario; A power prediction scenario generation module is used to generate a wind power prediction power sequence for each scenario based on the current prediction power sequence and the power prediction error sequence for each scenario; The power prediction error generating module includes: A joint distribution variable construction module is used to construct a joint distribution variable of the predicted power and power prediction error at adjacent historical moments based on the historical predicted power sequence and the historical measured power sequence; A clustering module is used to cluster the variables of the joint distribution of the predicted power and power prediction error at adjacent historical moments, and obtain the cluster centers of each category and the probability density function of the joint distribution of the predicted power and power prediction error at adjacent historical moments; A calculation module is used to calculate the power prediction error at the next moment under the conditions of the current predicted power at two given adjacent moments and the power prediction error at the previous moment in the two adjacent moments based on the current predicted power sequence, the cluster centers of each category and the joint distribution probability density function of the predicted power and power prediction error at the historical adjacent moments of each category.

7. The system according to claim 6, wherein: The computing module includes: A first calculation submodule is used to calculate the power prediction error at the first moment and the second moment in the power prediction error sequence based on the current predicted power sequence, the cluster centers of each category, and the joint distribution probability density function of the predicted power and power prediction error at historical adjacent moments of each category; A second calculation submodule is used to calculate, based on the current predicted power sequence, the cluster centers of each category, the joint distribution probability density function of the predicted power and power prediction error of each category at historical adjacent moments, and the power prediction error at the first moment and the second moment in the power prediction error sequence, the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence under the conditions of the current predicted power at two adjacent moments and the power prediction error at the previous moment between the two adjacent moments; The third calculation submodule is used to randomly sample the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence in sequence to obtain the power prediction error at the third moment and subsequent moments.

8. The system according to claim 7, wherein: The first calculation submodule includes: an initial high-dimensional vector construction unit, configured to construct a high-dimensional vector at a first adjacent moment based on the current predicted power sequence; the high-dimensional vector at the first adjacent moment being composed of the current predicted power at the first moment, the current predicted power at the second moment, an initial value set for the power prediction error at the first moment, and an initial value set for the power prediction error at the second moment; an initial high-dimensional vector classification unit, configured to determine the category to which the high-dimensional vector at the first adjacent moment belongs based on the cluster centers of the categories; The initial adjacent moment power prediction error calculation unit is used to randomly sample the predicted power and power prediction error joint distribution probability density function of the historical adjacent moments of the category to which the high-dimensional vector of the first adjacent moment belongs, and obtain the power prediction error of the first moment and the second moment in the power prediction error sequence.

9. The system according to claim 7, wherein: The second calculation submodule includes: Other high-dimensional vector construction units are used to sequentially generate high-dimensional vectors for each adjacent moment after the first adjacent moment based on the current predicted power sequence and the power prediction error at the second moment in the power prediction error sequence; the high-dimensional vectors for each adjacent moment after the first adjacent moment are composed of the current predicted power of the two adjacent moments, the power prediction error at the previous moment between the two adjacent moments, and the power prediction error at the next moment between the two adjacent moments. Initial values ​​are set; Other high-dimensional vector classification units are used to determine, based on the cluster centers of each category, the categories to which the high-dimensional vectors at each adjacent time after the first adjacent time belong; The power prediction error calculation unit at other moments is used to calculate the conditional probability distribution of the power prediction error at the third moment and subsequent moments in the power prediction error sequence based on the joint distribution probability density function of the predicted power and power prediction error of the historical adjacent moments of the category to which the high-dimensional vectors of each adjacent moment after the first adjacent moment belong.

Images