CN103530527A

CN103530527A - Wind power probability forecasting method based on numerical weather forecasting ensemble forecasting results

Info

Publication number: CN103530527A
Application number: CN201310524786.2A
Authority: CN
Inventors: 王铮; 王伟胜; 刘纯; 冯双磊; 王勃; 姜文玲; 赵艳青; 靳双龙; 胡菊; 王晓蓉; 张菲; 卢静; 车建峰; 马振强
Original assignee: State Grid Corp of China SGCC; China Electric Power Research Institute Co Ltd CEPRI; State Grid Anhui Electric Power Co Ltd; CLP Puri Zhangbei Wind Power Research and Test Ltd
Current assignee: State Grid Corp of China SGCC; China Electric Power Research Institute Co Ltd CEPRI; State Grid Anhui Electric Power Co Ltd; CLP Puri Zhangbei Wind Power Research and Test Ltd
Priority date: 2013-10-30
Filing date: 2013-10-30
Publication date: 2014-01-22

Abstract

The present invention provides a wind power probabilistic prediction method based on numerical weather forecast ensemble forecast results. Based on numerical weather forecast, the numerical weather forecast ensemble forecast technology provides basic input data for short-term wind power forecasting, and establishes for each ensemble member For the short-term forecasting model, multiple sets of forecast results are obtained. For the multiple sets of forecast results obtained, the ensemble forecast configuration characteristic division method and the forecast power level division method classify the prediction errors of different characteristics, and then use the non-parametric regression method to characterize the historical prediction of different characteristics The probability distribution of the error, and then obtain the future prediction error band under a certain level of confidence. The wind power probability prediction method provided by the present invention has a narrower error band interval under the same confidence level, and can effectively reduce the operating cost of the power grid under the same safety margin conditions for a power grid with large-scale wind power access , Improve the economy of power grid operation.

Description

A Probabilistic Prediction Method of Wind Power Based on Numerical Weather Prediction Ensemble Forecast Results

技术领域technical field

本发明涉及风电功率预测领域，具体涉及一种基于数值天气预报集合预报结果的风电功率概率预测方法。The invention relates to the field of wind power prediction, in particular to a wind power probability prediction method based on numerical weather forecast ensemble prediction results.

背景技术Background technique

风能是可再生能源的重要组成部分，风电是目前最成熟、最具规模开发和商业化发展前景的可再生能源发电方式之一。但风电与常规能源不同，具有很大的随机性、间歇性和不可控性，大规模的风电并入电网，对电网的规划建设、运行调度、分析控制、经济运行和电能质量等产生一定的影响。对风电场输出功率进行预测，是应对大规模风电接入电网的重要举措之一，风电功率预测能够为电网的调度运行提供技术支撑，增强系统的安全性和稳定性，同时还有益于风电场运行维护计划的制定。但风电功率具有很强的随机波动性，风的产生规律难以把握，导致风电预测误差较大，预测结果很难对电网调度计划的制定提供有效依据，因而在风电功率确定性预测基础上开展风电功率的概率预测技术研究，得出一定置信度下的预测结果误差带区间，可为大规模风电场并网的优化调度提供基础技术支撑，因而具有重要的实际应用价值。Wind energy is an important part of renewable energy, and wind power is currently one of the most mature, large-scale development and commercial development prospects of renewable energy power generation. However, wind power is different from conventional energy in that it is highly random, intermittent, and uncontrollable. Large-scale wind power is integrated into the power grid, which will have certain impact on the planning and construction, operation scheduling, analysis and control, economic operation and power quality of the power grid. Influence. Predicting the output power of wind farms is one of the important measures to deal with large-scale wind power access to the grid. Wind power prediction can provide technical support for the dispatching operation of the grid, enhance the security and stability of the system, and benefit wind farms. Formulation of operation and maintenance plan. However, wind power has strong random fluctuations, and the law of wind generation is difficult to grasp, resulting in large wind power prediction errors, and the prediction results are difficult to provide an effective basis for the formulation of grid dispatching plans. The research on the probabilistic prediction technology of power has obtained the error band interval of the prediction results under a certain degree of confidence, which can provide basic technical support for the optimal scheduling of large-scale wind farm grid-connected, so it has important practical application value.

当前风电功率的误差带概率预测主要采用确定的参数分布模型，如高斯分布、双曲分布以及贝塔分布等，拟合风电功率的历史预测误差分布，未识别不同特性的预测误差，所得到的概率预测误差带对所有预测结果的带宽相同，在相同置信度下宽度较大，不利于风电的经济调度运行。The current error band probability prediction of wind power mainly adopts certain parameter distribution models, such as Gaussian distribution, hyperbolic distribution, and beta distribution, etc., to fit the historical prediction error distribution of wind power, without identifying the prediction errors of different characteristics, the obtained probability The prediction error band has the same bandwidth for all prediction results, and the width is relatively large under the same confidence level, which is not conducive to the economic dispatch operation of wind power.

发明内容Contents of the invention

本发明针对现有技术的不足，提供一种基于数值天气预报集合预报结果的风电功率概率预测方法，包括：Aiming at the deficiencies of the prior art, the present invention provides a wind power probability prediction method based on numerical weather prediction ensemble prediction results, including:

步骤1，对数值天气预报集合预报的每个成员建立短期风电功率预测模型，分别输入数值天气预测结果后获得各个所述成员的风功率预测结果，得到天气预报集合预报的集合的风功率预测结果；Step 1. Establish a short-term wind power forecasting model for each member of the ensemble forecast of numerical weather, input the numerical weather forecast results respectively and obtain the forecast results of wind power of each member, and obtain the forecast result of wind power of the ensemble forecast of weather forecast ;

步骤2，根据所述集合中各成员的所述风电功率预测结果计算所述集合的二阶中心距，根据所述二阶中心距的设置的判别阀值识别各个所述集合的误差类型，所述集合的误差类型序号为i；Step 2, calculate the second-order center distance of the set according to the wind power prediction results of each member in the set, and identify the error type of each set according to the set discrimination threshold of the second-order center distance, so The error type sequence number of the set is i;

步骤3，根据所述集合的风功率预测结果将各个所述集合的功率水平进行划分，所述集合的功率水平序号为j；Step 3, divide the power levels of each of the sets according to the wind power prediction results of the sets, and the power level sequence number of the sets is j;

步骤4，计算误差类型为i功率水平为j的集合的相对误差的集合{e_ij}；所述相对误差的集合中各成员的相对误差

p_M为修正后的实际功率，S_op为装机容量；Step 4, calculate the relative error set {e _ij } of the set whose error type is i and the power level is j; the relative error of each member in the set of relative errors

p _M is the corrected actual power, S _op is the installed capacity;

步骤5，用核估计的方法得到所述集合{e_ij}中各个样本的概率密度函数；Step 5, obtain the probability density function of each sample in the set {e _ij } by means of kernel estimation;

步骤6，采用非参数拟合的方法拟合所述集合{e_ij}的概率密度分布，根据核回归理论得到所述集合{e_ij}的概率密度的拟合回归函数m_ij(e_ij)；Step 6, using a non-parametric fitting method to fit the probability density distribution of the set {e _ij }, and obtain the fitted regression function m _ij (e _ij ) of the probability density of the set {e _ij } according to the kernel regression theory ;

步骤7，对所述步骤6得到的所述概率密度的拟合回归结果进行回归校验；Step 7, performing regression check on the fitted regression result of the probability density obtained in the step 6;

步骤8，计算得到误差类型为i功率水平为j下置信水平等于1-α时的误差上限和误差下限

和

Step 8, calculate the error upper limit and error lower limit when the error type is i and the power level is j and the confidence level is equal to 1-α

and

步骤9，根据所述步骤8中得到的误差上下限计算得到置信度水平等于1-α的误差类型为i功率水平为j的功率预测结果p_ij的估计区间Δp_ij。Step 9: Calculate the estimated interval Δp _ij of the power prediction result p _ij whose confidence level is equal to 1-α and the error type is i and the power level is j according to the upper and lower limits of the error obtained in step 8.

本发明提供的第一优选实施例中：所述步骤1得到的所述天气预报集合预报的集合的风功率预测结果p为

In the first preferred embodiment provided by the present invention: the wind power prediction result p of the set of weather forecast sets obtained in step 1 is

p_m为第m个集合成员的风电功率预测结果，N表示集合成员的总个数。p _m is the wind power prediction result of the mth set member, and N represents the total number of set members.

本发明提供的第二优选实施例中：所述步骤2中，所述集合的二阶中心距

为各集合预测结果的平均值；In the second preferred embodiment provided by the present invention: in the step 2, the second-order center distance of the set

The average value of the predicted results for each set;

所述集合的误差类型分为准确、较准确和不准确3类，所述判别阀值为T₁和T₂：The error types of the set are divided into three categories: accurate, relatively accurate and inaccurate, and the discrimination thresholds are _T1 and _T2 :

b＜T₁时，集合的误差类型为“准确”，对应的误差类型序号i＝1；T₁≤b≤T₂时，集合的误差类型为“较准确”，对应的误差类型序号i＝2；b＞T₂时，集合的误差类型为“不准确”，对应的误差类型序号i＝3。When b<T ₁ , the error type of the set is "accurate", and the corresponding error type number i = 1; when T ₁ ≤ b ≤ _{T 2} , the error type of the set is "more accurate", and the corresponding error type number i= 2; when b>T ₂ , the error type of the set is "inaccurate", and the corresponding error type number i=3.

本发明提供的第三优选实施例中：所述步骤3中，所述集合的功率水平序号j的计算方法为：In the third preferred embodiment provided by the present invention: in the step 3, the calculation method of the power level number j of the set is:

j为正整数；

j is a positive integer;

η为功率水平划分系数；p_n为装机容量，单位为MW；

η is the power level division coefficient; p _n is the installed capacity, the unit is MW;

本发明提供的第四优选实施例中：所述步骤5中，用核估计的方法得到的所述集合{e_ij}中各个样本的概率密度函数为：In the fourth preferred embodiment provided by the present invention: in the step 5, the probability density function of each sample in the set {e _ij } obtained by means of kernel estimation is:

${f f}_{ij ij} (({e e}_{ijk ijk})) = = \frac{11}{n no {δ δ}_{ij ij}} {Σ Σ}_{M m = = 11}^{n no} {K K}_{11} ((\frac{{e e}_{ijk ijk} - - {e e}_{ijM i}}{{δ δ}_{ij ij}}));;$

n为集合{e_ij}中的样本总数；

为核估计的核函数，该核函数K₁()取均匀核，δ_ij为窗宽；k表示窗宽δ_ij中的样本的序号，f_ij(e_ijk)为窗宽δ_ij中任意一个样本e_ijk的概率密度函数；1≤M≤n。n is the total number of samples in the set {e _ij };

is the kernel function of kernel estimation, the kernel function K ₁ () takes a uniform kernel, δ _ij is the window width; k represents the serial number of the sample in the window width δ _ij , f _ij (e _ijk ) is any one of the window width δ _ij Probability density function of sample e _ijk ; 1≤M≤n.

本发明提供的第五优选实施例中：所述步骤6中根据核回归理论得到误差分布集合{e_ij}的概率密度的拟合回归函数m_ij(e_ij)为：In the fifth preferred embodiment provided by the present invention: in the step 6, the fitted regression function m _ij (e _ij ) of the probability density of the error distribution set {e _ij } obtained according to the kernel regression theory is:

${m m}_{ij ij} (({e e}_{ij ij})) = = [[{Σ Σ}_{k k = = 11}^{n no} K K ((\frac{{e e}_{ij ij} - - {e e}_{ijk ijk}}{{h h}_{ij ij}})) {f f}_{ij ij} (({e e}_{ijk ijk}))]] / / {Σ Σ}_{k k = = 11}^{n no} K K ((\frac{{e e}_{ij ij} - - {e e}_{ijk ijk}}{{h h}_{ij ij}}));;$

其中，

为核函数标准正态核：in,

Standard normal kernel for the kernel function:

$K K ((\frac{{e e}_{ij ij} - - {e e}_{ijk ijk}}{{h h}_{ij ij}})) = = {((22 π π))}^{- - \frac{11}{22}} exp exp ((- - \frac{11}{22} {((\frac{{e e}_{ij ij} - - {e e}_{ijk ijk}}{{h h}_{ij ij}}))}^{22}));;$

h_ij为核回归窗宽。h _ij is the kernel regression window width.

本发明提供的第六优选实施例中：对所述步骤6得到的误差分布的概率密度的拟合回归结果进行回归校验，得到满足校验结果的最大的核回归窗宽h_ij值：In the sixth preferred embodiment provided by the present invention: regression verification is performed on the fitting regression result of the probability density of the error distribution obtained in step 6, and the maximum kernel regression window width h _ij value satisfying the verification result is obtained:

校验过程包括：The verification process includes:

以i误差类型j功率水平下的所有误差U_ij为母体，划分为有限多项的离散分布，设A₁,...,A_l为不同误差事件，满足：

其中，l的取值与h_ij有关；Taking all errors U _ij of i error type j power level as the parent body, they are divided into discrete distributions of finite multinomials. Let A ₁ ,...,A _l be different error events, satisfying:

Among them, the value of l is related to h _ij ;

根据所述拟合回归函数m_ij(x)可得到不同误差下的概率：According to the fitting regression function m _ij (x), the probability under different errors can be obtained:

y_ijk＝m_ij(A_k)·h_ij,k＝1,...,l；y _ijk = m _ij (A _k ) h _ij ,k=1,...,l;

得到A_k误差下的抽样频数m_ijk满足

n_ij为误差样本{e_ij}的容量；Get the sampling frequency m _ijk under A _k error to satisfy

n _ij is the capacity of the error sample {e _ij };

考察样本的实际频数m_ijk对理论频数n_ij·y_ijk偏差的加权平方和

当n_ij较大时，χ²近似的服从自由度为l-1的χ²分布；Investigate the weighted sum of squares of the deviation between the actual frequency m _ijk of the sample and the theoretical frequency n _ij ·y _ijk

When n _ij is large, χ ² approximately obeys the χ ² distribution with a degree of freedom of l-1;

给定显著性水平α，若

则认为拟合结果与实际情况是相符的，具有可信性，即校验通过，否则不可信，校验不通过；Given a significance level α, if

Then it is considered that the fitting result is consistent with the actual situation and has credibility, that is, the verification is passed, otherwise it is not credible and the verification fails;

其中

的值由χ²分布上侧分位表查得。in

The value of is obtained from the upper quantile table of the χ ² distribution.

本发明提供的第七优选实施例中：所述步骤8中包括：In the seventh preferred embodiment provided by the present invention: said step 8 includes:

根据所述拟合回归函数m_ij(e_ij)得到总体分布函数：Obtain the overall distribution function according to the fitting regression function m _ij (e _ij ):

${F f}_{ij ij} ((e e)) = = {&Integral; &Integral;}_{- - 11}^{e e} {m m}_{ij ij} (({e e}_{ij ij})) \cdot &Center Dot; {h h}_{ij ij} {de de}_{ij ij} e e &Element; &Element; [[- - 1,1 1,1]];;$

其中，e为功率预测误差取值，m_ij(e_ij)应该满足

Among them, e is the power prediction error value, m _ij (e _ij ) should satisfy

满足 $F_{ij} (β_{ij}^{u}) - F_{ij} (β_{ij}^{l}) = 1 - α$ 并且

的值最小的

和

即为误差类型为i功率水平为j下置信水平等于1-α的误差上限和误差下限。satisfy

f_{ij} (β_{ij}^{u}) - f_{ij} (β_{ij}^{l}) = 1 - α

and

The value of the smallest

and

That is, the error type is i and the power level is j, and the confidence level is equal to the error upper limit and error lower limit of 1-α.

本发明提供的第八优选实施例中：所述步骤9中的所述的估计区间Δp_ij为 $Δ p_{ij} = [p_{ij}^{lL}, p_{ij}^{uL}] :$ In the eighth preferred embodiment provided by the present invention: the estimation interval Δp _ij in the step 9 is $Δ p_{ij} = [p_{ij}^{L}, p_{ij}^{uL}] :$

${p p}_{ij ij}^{uL uL} = = \{\begin{matrix} {p p}_{ij ij} - - {β β}_{ij ij}^{lL L} \cdot \cdot {S S}_{op op} & {p p}_{ij ij}^{uL uL} \leq \leq {S S}_{op op} \\ {S S}_{op op} & {p p}_{ij ij}^{uL uL} > > {S S}_{op op} \end{matrix};;$ ${p p}_{ij ij}^{lL L} = = \{\begin{matrix} {p p}_{ij ij} - - {β β}_{ij ij}^{uL uL} \cdot \cdot {S S}_{op op} & {p p}_{ij ij}^{lL L} &GreaterEqual; &Greater Equal; 00 \\ 00 & {p p}_{ij ij}^{lL L} < < 00 \end{matrix} . .$

与现有技术比，本发明提供的具有以下优异效果：Compared with the prior art, the present invention provides the following excellent effects:

1、本发明提供的一种基于数值天气预报集合预报结果的风电功率概率预测方法，在大量分析和研究实际风电场数据基础上，发现短期风电功率预测误差的主要来源是数值天气预报，从而形成以数值天气预报为基础，通过数值天气预报集合预报技术为短期风电功率预测提供了基础输入数据，针对每一个集合成员建立短期预测模型，得到多组预测结果，研究风电功率概率预测，在相同置信度水平下，误差带区间更窄，对于含大规模风电接入的电网，在满足相同安全裕度条件下，采用本专利技术所得到的风电功率预测误差带，能够有效减少电网运行成本，提高电网运行的经济型。1. The present invention provides a wind power probabilistic prediction method based on numerical weather forecast ensemble forecast results. On the basis of a large number of analysis and research on actual wind farm data, it is found that the main source of short-term wind power prediction error is numerical weather forecast, thus forming Based on numerical weather prediction, the numerical weather prediction ensemble prediction technology provides basic input data for short-term wind power prediction, establishes a short-term prediction model for each ensemble member, obtains multiple sets of prediction results, and studies wind power probability prediction. Under the high degree level, the error band interval is narrower. For the power grid with large-scale wind power access, under the same safety margin, the wind power prediction error band obtained by using this patented technology can effectively reduce the operating cost of the power grid and improve Economical grid operation.

2、从风属于气象物理学研究范畴，具有其内在的本质规律属性出发，提出通过不同特性预测误差的分类识别来提高风电功率概率预测精准度的研究方法，采用集合预报组态特性划分方法和预测功率水平划分方法对不同特性预测误差进行分类，有效提高了风电功率概率预测的精准度。2. Starting from the fact that wind belongs to the research category of meteorological physics and has its inherent essential law attributes, a research method is proposed to improve the accuracy of wind power probability prediction by classifying and identifying prediction errors of different characteristics. The prediction power level division method classifies the prediction errors of different characteristics, which effectively improves the accuracy of wind power probability prediction.

3、采用非参数回归方法表征不同特性历史预测误差的概率分布，进而得出一定置信度水平下未来的预测误差带，达到通过历史预测误差特性的识别分析，把握未来预测误差特性的目的。3. Use the non-parametric regression method to characterize the probability distribution of historical prediction errors of different characteristics, and then obtain the future prediction error band under a certain level of confidence, so as to achieve the purpose of grasping the characteristics of future prediction errors through the identification and analysis of historical prediction error characteristics.

附图说明Description of drawings

如图1所示为本发明提供的一种基于数值天气预报集合预报结果的风电功率概率预测方法的流程图。FIG. 1 is a flow chart of a wind power probabilistic prediction method based on numerical weather prediction ensemble prediction results provided by the present invention.

具体实施方式Detailed ways

下面根据附图对本发明的具体实施方式作进一步详细说明。The specific implementation manner of the present invention will be described in further detail below according to the accompanying drawings.

本发明提供一种基于数值天气预报集合预报结果的风电功率概率预测方法，其流程图如图1所示，由图1可知，该方法包括：The present invention provides a method for probabilistic forecasting of wind power based on numerical weather forecast ensemble forecast results, the flow chart of which is shown in Figure 1, as can be seen from Figure 1, the method includes:

步骤1，对数值天气预报集合预报的每个成员建立短期风电功率预测模型，分别输入数值天气预测结果后获得各个成员的风功率预测结果，得到天气预报集合预报的集合的风功率预测结果。Step 1: Establish a short-term wind power prediction model for each member of the ensemble of numerical weather forecasts, input the numerical weather prediction results respectively to obtain the wind power prediction results of each member, and obtain the wind power prediction results of the ensemble of weather forecasts.

步骤2，根据集合中各成员的风电功率预测结果计算该集合的二阶中心距，根据该二阶中心距的设置的判别阀值识别各个集合的误差类型，将各个集合的误差类型分为三种，即集合的误差类型序号i＝1，2，3。Step 2, calculate the second-order center distance of the set according to the wind power prediction results of each member in the set, identify the error type of each set according to the discriminant threshold set by the second-order center distance, and divide the error types of each set into three types Types, that is, the sequence numbers of the error types of the set i=1, 2, 3.

步骤3，根据集合的风功率预测结果p将各个集合的功率水平进行划分，集合的功率水平序号为j。Step 3: Divide the power level of each set according to the wind power prediction result p of the set, and the power level sequence number of the set is j.

步骤4，对集合进行误差类型划分和功率水平划分后，计算误差类型为i功率水平为j的集合的相对误差的集合{e_ij}。相对误差的集合中各成员的相对误差

p_M为修正后的实际功率，S_op为装机容量。Step 4: After dividing the set by error type and power level, calculate the relative error set {e _ij } of the set whose error type is i and power level is j. The relative error of each member in the set of relative errors

p _M is the corrected actual power, S _op is the installed capacity.

步骤5，用核估计的方法得到{e_ij}中各个样本的概率密度函数。Step 5. Obtain the probability density function of each sample in {e _ij } by means of kernel estimation.

步骤6，采用非参数拟合的方法拟合集合{e_ij}的概率密度分布，根据核回归理论得到误差分布集合{e_ij}的概率密度的拟合回归函数m_ij(e_ij)。Step 6: Fit the probability density distribution of the set {e _ij } by non-parametric fitting method, and obtain the fitting regression function m _ij (e _ij ) of the probability density of the error distribution set {e _ij } according to the kernel regression theory.

步骤7，对步骤6得到的误差分布的概率密度的拟合回归结果进行回归校验。Step 7, perform regression verification on the fitting regression result of the probability density of the error distribution obtained in step 6.

和

and

步骤9，根据步骤8得到的误差上下限计算得到置信度水平等于1-α的误差类型为i功率水平为j的功率预测结果p_ij的估计区间Δp_ij。Step 9: Calculate the upper and lower limits of the error obtained in step 8 to obtain the estimated interval Δp _ij of the power prediction result p _ij whose confidence level is equal to 1-α and whose error type is i and whose power level is j.

进一步的，步骤1中，对数值天气预报集合预报的每个成员建立短期风电功率预测模型，分别输入数值天气预测结果后获得各个成员的风功率预测结果，得到天气预报集合预报的集合的风功率预测结果p为

Further, in step 1, a short-term wind power prediction model is established for each member of the numerical weather forecast ensemble forecast, and the wind power forecast results of each member are obtained after inputting the numerical weather prediction results respectively, and the wind power of the ensemble forecast of the weather forecast is obtained The predicted result p is

式中，p_m为第m个集合成员的风电功率预测结果，N表示集合成员的总个数。In the formula, p _m is the wind power prediction result of the mth ensemble member, and N represents the total number of ensemble members.

步骤2中，由于目前认知的有限性，集合预测中的各成员均无法准确把握天气的实际变化情况，但不同的初始场和参数化方案会对混沌特性较大的天气过程做出差别较大的反应，因而可利用各集合成员的组态特性把握不同特性的天气过程，在风电功率预测中表现为不同特性的预测误差，从而实现预测误差类型的识别划分。In step 2, due to the limited cognition at present, each member in the ensemble forecast cannot accurately grasp the actual change of the weather, but different initial fields and parameterization schemes will make different comparisons for weather processes with large chaotic characteristics. Therefore, the configuration characteristics of each set member can be used to grasp the weather process with different characteristics, which is manifested as prediction errors with different characteristics in wind power prediction, so as to realize the identification and division of prediction error types.

根据大量的分析研究发现，基于集合成员的组态特性可将预测结果大体分为准确、较准确和不准确3类，不同预测结果类型可根据集合预报结果的二阶中心距进行判定。According to a large number of analysis and research, based on the configuration characteristics of ensemble members, the prediction results can be roughly divided into three categories: accurate, relatively accurate and inaccurate. Different types of prediction results can be judged according to the second-order center distance of the ensemble prediction results.

集合中各成员的风电功率预测结果p_m计算该集合的二阶中心距

为各集合预测结果的平均值。Calculate the second-order center distance of the set from the wind power prediction results p _m of each member in the set

The average of the predicted results for each set.

假设“准确”与“较准确”的区分阀值为T₁，“较准确”与“不准确”的区分阀值为T₂。则b＜T₁时，集合的误差类型为“准确”，对应的误差类型序号i＝1；T₁≤b≤T₂时，集合的误差类型为“较准确”，对应的误差类型序号i＝2；b＞T₂时，集合的误差类型为“不准确”，对应的误差类型序号i＝3。Assume that the threshold for distinguishing "accurate" and "more accurate" is T ₁ , and the threshold for distinguishing "more accurate" and "inaccurate" is T ₂ . Then when b<T ₁ , the error type of the set is "accurate", and the corresponding error type number i = 1; when T ₁ ≤ b ≤ _{T 2} , the error type of the set is "more accurate", and the corresponding error type number i =2; when b>T ₂ , the error type of the set is "inaccurate", and the corresponding error type number i=3.

步骤3中，将数值天气预报输出的气象数据转化为风电场预测功率，但风电场功率曲线具有非线性特性，功率曲线在功率两端对预测的不确定性起抑制作用，但误差空间被放大，在数值天气预报数据存在误差情况下，为了使整体预测误差最小，预测数据将被压缩至误差空间相对较小的中段部分，以防极端误差出现，使得低风速和高风速水平下，预测误差的分布范围相对较小，不确定性较低，但容易出现极端误差；在中等风速水平下，误差分布范围较大，不确定性增加。因而不同预测功率水平，预测误差的特性不同。In step 3, the meteorological data output by the numerical weather prediction is converted into the predicted power of the wind farm, but the power curve of the wind farm has nonlinear characteristics, and the power curve suppresses the uncertainty of the prediction at both ends of the power, but the error space is enlarged , in the case of errors in the numerical weather prediction data, in order to minimize the overall prediction error, the prediction data will be compressed to the middle part of the relatively small error space to prevent extreme errors, so that the prediction error at low and high wind speeds The distribution range of is relatively small, and the uncertainty is low, but extreme errors are prone to occur; at the medium wind speed level, the error distribution range is large, and the uncertainty increases. Therefore, the characteristics of the prediction error are different for different prediction power levels.

根据集合的风功率预测结果p将各个集合的功率水平进行划分，其中，集合的功率水平序号j为：The power level of each set is divided according to the wind power prediction result p of the set, where the power level sequence number j of the set is:

j为正整数。

j is a positive integer.

η为功率水平划分系数；p_n为装机容量，单位为MW；

即将功率水平划分为以下

个水平范围：

[0, η p_{n}], (η p_{n}, 2 η p_{n}] . . . . . . (η p_{n} \times (\frac{1}{η} - 1), η p_{n} \times \frac{1}{η}],

对应的功率水平序号j为

That is, the power levels are divided into the following

horizontal range:

[0, η p_{no}], (η p_{no}, 2 η p_{no}] . . . . . . (η p_{no} \times (\frac{1}{η} - 1), η p_{no} \times \frac{1}{η}],

The corresponding power level number j is

功率水平越细化，越能掌握误差的内在性质，但在数据总体一定情况下，功率水平细化越多，每个细化样本中的数据量相应减小，可能导致数据样本不能够反映相应水平下误差的真实分布，两者产生矛盾。因而功率水平的细化程度，应根据具体数据样本和总体误差情况确定。The more refined the power level is, the better the internal nature of the error can be grasped. However, in the case of a certain data population, the more the power level is refined, the amount of data in each refined sample will be correspondingly reduced, which may cause the data sample to fail to reflect the corresponding error. The true distribution of the error at the lower level, the two produce a contradiction. Therefore, the degree of refinement of the power level should be determined according to the specific data sample and the overall error situation.

步骤4，对集合进行误差类型划分和功率水平划分后，计算相对误差集{e_ij}。{e_ij}为误差类型为i功率水平为j的集合的相对误差的集合，相对误差的集合中各成员的相对误差

p_M为修正后的实际功率，S_op为装机容量。Step 4, after dividing the set by error type and power level, calculate the relative error set {e _ij }. {e _ij } is the relative error set of the set whose error type is i and the power level is j, and the relative error of each member in the set of relative errors

p _M is the corrected actual power, S _op is the installed capacity.

步骤5，用核估计的方法得到{e_ij}中各个样本的概率密度函数为：Step 5, use the method of kernel estimation to obtain the probability density function of each sample in {e _ij } as follows:

n为集合{e_ij}中的样本总数；

为核估计的核函数，该核函数K₁()可以取均匀核，δ_ij为窗宽；k表示窗宽δ_ij中的样本的序号，f_ij(e_ijk)为窗宽δ_ij中任意一个样本e_ijk的概率密度函数；1≤M≤n。n is the total number of samples in the set {e _ij };

is the kernel function of kernel estimation, the kernel function K ₁ () can take a uniform kernel, δ _ij is the window width; k represents the serial number of the sample in the window width δ _ij , f _ij (e _ijk ) is any of the window width δ _ij Probability density function of a sample e _ijk ; 1≤M≤n.

步骤6，采用非参数拟合的方法拟合集合{e_ij}的概率密度分布，根据核回归理论得到误差分布集合{e_ij}的概率密度的拟合回归函数m_ij(e_ij)：Step 6: Fit the probability density distribution of the set {e _ij } by non-parametric fitting method, and obtain the fitting regression function m _ij (e _ij ) of the probability density of the error distribution set {e _ij } according to the kernel regression theory:

其中，

为核回归的核函数，h_ij为核回归窗宽，该核函数K()可以取标准正态核：in,

is the kernel function of kernel regression, h _ij is the window width of kernel regression, and the kernel function K() can take the standard normal kernel:

$K K ((\frac{{e e}_{ij ij} - - {e e}_{ijk ijk}}{{h h}_{ij ij}})) = = {((22 π π))}^{- - \frac{11}{22}} exp exp ((- - \frac{11}{22} {((\frac{{e e}_{ij ij} - - {e e}_{ijk ijk}}{{h h}_{ij ij}}))}^{22})) . .$

步骤7，对步骤6得到的误差分布的概率密度的拟合回归结果进行回归校验，得到满足校验结果的最大的核回归窗宽h_ij值。Step 7: Perform regression check on the fitted regression result of the probability density of the error distribution obtained in Step 6, and obtain the maximum kernel regression window width h _ij value that satisfies the check result.

采用非参数回归拟合得到了误差的概率分布，但需对分布拟合结果进行校验，满足校验结果后才能采用，本发明提供的确定方法是在满足检验条件的前提下h_ij尽可能取较大值。本发明采用卡方校验法对拟合结果进行校验，校验过程包括：The probability distribution of the error is obtained by non-parametric regression fitting, but the distribution fitting result needs to be verified, and it can be adopted after the verification result is satisfied. The determination method provided by the present invention is to satisfy the test condition as _far as possible. Take the larger value. The present invention adopts chi-square verification method to verify the fitting result, and the verification process includes:

以i误差类型j功率水平下的所有误差U_ij为母体，其可以划分为有限多项的离散分布。设A₁,...,A_l为不同误差事件，满足：

Taking all errors U _ij under i error type j power level as parent, it can be divided into discrete distribution of finite multinomials. Let A ₁ ,...,A _l be different error events, satisfying:

其中，l的取值与h_ij有关。Among them, the value of l is related to h _ij .

根据回归函数m_ij(x)可得到不同误差下的概率：According to the regression function m _ij (x), the probability under different errors can be obtained:

y_ijk＝m_ij(A_k)·h_ij,k＝1,...,l。y _ijk = m _ij (A _k ) · h _ij , k=1, . . . , l.

{e_ij}为i误差类型j功率水平下的误差样本，假设n_ij为误差样本{e_ij}的容量。于是可以得到A_k误差下的抽样频数m_ijk满足{e _ij } is the error sample under i error type j power level, assuming n _ij is the capacity of error sample {e _ij }. Then we can get the sampling frequency m _ijk under A _k error to satisfy

${Σ Σ}_{k k = = 11}^{l l} {m m}_{ijk ijk} = = {n no}_{ij ij},, k k = = 11,, . . . . . .,, l l . .$

如果回归拟合结果与实际情况相符，那么位于误差A_k下的抽样频数m_ijk应该接近于n_ij·y_ijk。If the regression fitting result is consistent with the actual situation, then the sampling frequency m _ijk under the error A _k should be close to n _ij ·y _ijk .

统计量χ²值的大小反映了子样实际频数分布对理论频数分布的拟合程度。当n_ij较大时，χ²近似的服从自由度为l-1的χ²分布。Investigate the weighted sum of squares of the deviation between the actual frequency m _ijk of the sample and the theoretical frequency n _ij ·y _ijk

The size of the statistic χ ² reflects the fitting degree of the actual frequency distribution of the sample to the theoretical frequency distribution. When n _ij is large, χ ² approximately obeys the χ ² distribution with the degree of freedom l-1.

给定显著性水平α，若则认为拟合结果与实际情况是相符的，具有可信性，否则不可信。其中

的值可由χ²分布上侧分位表查得。Given a significance level α, if It is considered that the fitting result is consistent with the actual situation and is credible, otherwise it is not credible. in

The value of can be found from the upper quantile table of the χ ² distribution.

步骤8中，当拟合分布结果满足检验要求后，可进一步建立预测结果的误差带，否则需调整参数重新进行拟合，包括：In step 8, when the fitting distribution results meet the inspection requirements, the error band of the prediction result can be further established, otherwise, the parameters need to be adjusted to re-fit, including:

根据拟合回归函数m_ij(e_ij)得到总体分布函数：According to the fitting regression function m _ij (e _ij ), the overall distribution function is obtained:

${F f}_{ij ij} ((e e)) = = {&Integral; &Integral;}_{- - 11}^{e e} {m m}_{ij ij} (({e e}_{ij ij})) \cdot &Center Dot; {h h}_{ij ij} {de de}_{ij ij} e e &Element; &Element; [[- - 1,1 1,1]] . .$

其中，e为功率预测误差取值，m_ij(e_ij)应该满足

满足

的

和

有无数组，但间距最小的只有一组，在此称为带宽最小原则。即满足

F_{ij} (β_{ij}^{u}) - F_{ij} (β_{ij}^{l}) = 1 - α

并且

的值最小时，

和

即是误差类型为i功率水平为j下置信水平等于1-α的误差上限和误差下限。satisfy

of

and

There are infinite groups, but only one group with the smallest spacing, which is called the principle of minimum bandwidth. That is satisfied

f_{ij} (β_{ij}^{u}) - f_{ij} (β_{ij}^{l}) = 1 - α

and

When the value of is the smallest,

and

That is, when the error type is i and the power level is j, the confidence level is equal to the error upper limit and error lower limit of 1-α.

步骤9，根据步骤8得到的误差上下限计算得到置信度水平等于1-α的误差类型为i功率水平为j的功率预测结果p_ij的估计区间Δp_ij： Step 9: Calculate the upper and lower limits of the error obtained in step 8 to obtain the estimated interval Δp _ij of the power prediction result p _ij whose confidence level is equal to 1-α, the error type is i, and the power level is j:

具体的，当确立误差上限和下限之后，需要将误差区间转化成预测功率区间，根据 $e = \frac{P - P_{M}}{S_{op}}$ 可得： $\{\begin{matrix} p_{ij}^{uL} = p_{ij} - β_{ij}^{lL} \cdot S_{op} \\ p_{ij}^{lL} = p_{ij} - β_{ij}^{uL} \cdot S_{op} \end{matrix} .$ Specifically, after establishing the upper and lower limits of the error, it is necessary to convert the error interval into a predicted power interval, according to $e = \frac{P - P_{m}}{S_{op}}$ Available: $\{\begin{matrix} p_{ij}^{uL} = p_{ij} - β_{ij}^{L} &Center Dot; S_{op} \\ p_{ij}^{L} = p_{ij} - β_{ij}^{uL} &Center Dot; S_{op} \end{matrix} .$

式中，p_ij为i误差类型j功率水平下的预测功率。功率上限

可能超出装机容量S_op的限制，但实际上功率上限不可能超过装机容量，因而In the formula, p _ij is the predicted power under the i error type j power level. power cap

may exceed the limit of the installed capacity S _op , but in fact the power upper limit cannot exceed the installed capacity, so

${p p}_{ij ij}^{uL uL} = = \{\begin{matrix} {p p}_{ij ij} - - {β β}_{ij ij}^{lL L} \cdot &Center Dot; {S S}_{op op} & {p p}_{ij ij}^{uL uL} \leq \leq {S S}_{op op} \\ {S S}_{op op} & {p p}_{ij ij}^{uL uL} > > {S S}_{op op} \end{matrix} . .$

同样，功率下限

可能变为负数，于是Likewise, the power lower limit

may become negative, so

${p p}_{ij ij}^{lL L} = = \{\begin{matrix} {p p}_{ij ij} - - {β β}_{ij ij}^{uL uL} \cdot &Center Dot; {S S}_{op op} & {p p}_{ij ij}^{lL L} &GreaterEqual; &Greater Equal; 00 \\ 00 & {p p}_{ij ij}^{lL L} < < 00 \end{matrix} . .$

经过实际风电场测试表明，该方法能够有效得出风电功率预测结果的误差带，且通过与当前主要概率预测方法对比发现，在相同置信度水平下本文所提方法能够获得更高精准度的预测误差带。The actual wind farm test shows that the method can effectively obtain the error band of the wind power prediction results, and by comparing with the current main probability prediction methods, it is found that the method proposed in this paper can obtain higher accuracy prediction under the same confidence level error band.

最后应当说明的是：以上实施例仅用以说明本发明的技术方案而非对其限制，尽管参照上述实施例对本发明进行了详细的说明，所属领域的普通技术人员应当理解：依然可以对本发明的具体实施方式进行修改或者等同替换，而未脱离本发明精神和范围的任何修改或者等同替换，其均应涵盖在本发明的权利要求范围当中。Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the present invention can still be Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the scope of the claims of the present invention.

Claims

1. A wind power probability prediction method based on numerical weather forecast ensemble forecast results is characterized by comprising the following steps:

step 1, establishing a short-term wind power prediction model for each member of the numerical weather forecast ensemble, and respectively inputting numerical weather prediction results to obtain wind power prediction results of each member so as to obtain wind power prediction results of the ensemble of the weather forecast ensemble;

step 2, calculating second-order center distances of the sets according to the wind power prediction results of all members in the sets, and identifying error types of all the sets according to set judgment thresholds of the second-order center distances, wherein the error type serial numbers of the sets are i;

step 3, dividing the power level of each set according to the wind power prediction result of the set, wherein the power level serial number of the set is j;

step 4, calculating a set { e) of relative errors of a set with the error type i and the power level j_ij}; relative error of each member of the set of relative errors

p_MFor the corrected actual power, S_opIs installed capacity;

step 5, obtaining the set { e ] by using a kernel estimation method_ij-probability density function of each sample in the };

step 6, fitting the set { e ] by adopting a non-parameter fitting method_ijObtaining the set { e } according to a kernel regression theory based on the probability density distribution of the set { e }, wherein the set is a set of the set { e } and the set is a set of the set { e } and the set is a set of the set_ijFitted regression function m of probability density of_ij(e_ij)；

7, performing regression verification on the fitting regression result of the probability density obtained in the step 6;

step 8, calculating an error upper limit and an error lower limit when the power level of the error type i is j and the confidence level is equal to 1-alpha

And

step 9, calculating to obtain a power prediction result p with the confidence level equal to 1-alpha and the error type i with the power level j according to the upper and lower error limits obtained in the step 8_ijIs estimated by the interval Δ p_ij。

2. The method of claim 1, wherein the first and second light sources are selected from the group consisting of a red light source, a green light source, and a blue light source,wherein the wind power prediction result p of the set of weather forecast ensemble forecast obtained in the step 1 is

p_mAnd N represents the total number of the set members as the wind power prediction result of the mth set member.

3. The method of claim 2, wherein in step 2, the second order center-to-center distance of the set The average of the predicted results for each set;

the error types of the set are divided into 3 types of accurate, more accurate and inaccurate, and the discrimination threshold is T₁And T₂：

b＜T₁Then, the error type of the set is accurate, and the corresponding error type serial number i is 1; t is₁≤b≤T₂Then, the error type of the set is more accurate, and the corresponding error type serial number i is 2; b > T₂In the case of a set, the error type is "inaccurate", and the corresponding error type number i is 3.

4. The method of claim 2, wherein in step 3, the power level index j of the set is calculated by:

j is a positive integer;

eta is a power level division coefficient; p is a radical of_nInstalled capacity in MW;

5. the method of claim 1, wherein in step 5, the set { e ] is obtained by a kernel estimation method_ijThe probability density function for each sample in the } is:

n is the set { e }_ijTotal number of samples in (j);

kernel function for kernel estimation, said kernel function K₁() Taking a uniform kernel, delta_ijIs the window width; k represents the window width delta_ijNumber of samples in (1), f_ij(e_ijk) Is window width delta_ijAny one of the samples e_ijkA probability density function of; m is more than or equal to 1 and less than or equal to n.

6. The method of claim 5, wherein the error distribution set { e } is obtained in step 6 according to kernel regression theory_ijFitted regression function m of probability density of_ij(e_ij) Comprises the following steps:

wherein,standard normal kernels for kernel function:

h_ijthe kernel regression window width.

7. The method according to claim 6, wherein the fitting regression result of the probability density of the error distribution obtained in the step 6 is subjected to regression check to obtain a maximum kernel regression window width h satisfying the check result_ijThe value:

the checking process comprises the following steps:

all errors U at power level of i error type j_ijFor the precursor, divided into a discrete distribution of finite elements, let A₁,...,A_lFor different error events, the following are satisfied:

wherein, the value of l and h_ij(ii) related;

according to the fitted regression function m_ij(x) The probabilities under different errors can be obtained:

y_ijk＝m_ij(A_k)·h_ij,k＝1,...,l；

to obtain A_kSampling frequency m under error_ijkSatisfy the requirement ofn_ijIs an error sample e_ijThe capacity of the device;

actual frequency m of investigation samples_ijkFor the theoretical frequency n_ij·y_ijkWeighted sum of squares of deviationsWhen n is_ijWhen it is bigger, x²Approximate compliance chi with degree of freedom l-1²Distributing;

giving a significance level α, if

The fitting result is considered to be consistent with the actual situation and has credibility, namely, the verification is passed, otherwise, the verification is not passed;

wherein

Has a value of x²And (6) looking up a distribution upper side division table.

8. The method of claim 6, wherein step 8 comprises:

according to the fitted regression function m_ij(e_ij) The overall distribution function is obtained:

<math> <mrow> <msub> <mi>F</mi> <mi>ij</mi> </msub> <mrow> <mo>(</mo> <mi>e</mi> <mo>)</mo> </mrow> <mo>=</mo> <msubsup> <mo>&Integral;</mo> <mrow> <mo>-</mo> <mn>1</mn> </mrow> <mi>e</mi> </msubsup> <msub> <mi>m</mi> <mi>ij</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>e</mi> <mi>ij</mi> </msub> <mo>)</mo> </mrow> <mo>·</mo> <msub> <mi>h</mi> <mi>ij</mi> </msub> <msub> <mi>de</mi> <mi>ij</mi> </msub> <mi>e</mi> <mo>&Element;</mo> <mo>[</mo> <mo>-</mo> <mn>1,1</mn> <mo>]</mo> <mo>;</mo> </mrow> </math>

wherein e is a power prediction error value, m_ij(e_ij) Should satisfy

Satisfy the requirement of

And is

Of minimum value

And

i.e. an upper error limit and a lower error limit with a confidence level equal to 1-alpha with an error type i and a power level j.

9. The method of claim 6, wherein said estimation interval Δ p in said step 9_ijIs composed of

<math> <mrow> <msubsup> <mi>p</mi> <mi>ij</mi> <mi>lL</mi> </msubsup> <mo>=</mo> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <msub> <mi>p</mi> <mi>ij</mi> </msub> <mo>-</mo> <msubsup> <mi>β</mi> <mi>ij</mi> <mi>uL</mi> </msubsup> <mo>·</mo> <msub> <mi>S</mi> <mi>op</mi> </msub> </mtd> <mtd> <msubsup> <mi>p</mi> <mi>ij</mi> <mi>lL</mi> </msubsup> <mo>&GreaterEqual;</mo> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <msubsup> <mi>p</mi> <mi>ij</mi> <mi>lL</mi> </msubsup> <mo><</mo> <mn>0</mn> </mtd> </mtr> </mtable> </mfenced> <mo>.</mo> </mrow> </math>