CN103984986A

CN103984986A - Method for correcting wind power ultra-short-period prediction of self-learning ARMA model in real time

Info

Publication number: CN103984986A
Application number: CN201410186902.9A
Authority: CN
Inventors: 汪宁渤; 路亮; 赵龙; 张金平; 黄蓉
Original assignee: State Grid Corp of China SGCC; State Grid Gansu Electric Power Co Ltd; Wind Power Technology Center of Gansu Electric Power Co Ltd
Current assignee: State Grid Corp of China SGCC; State Grid Gansu Electric Power Co Ltd; Wind Power Technology Center of Gansu Electric Power Co Ltd
Priority date: 2014-05-06
Filing date: 2014-05-06
Publication date: 2014-08-13
Anticipated expiration: 2034-05-06
Also published as: CN103984986B

Abstract

The invention discloses a method for correcting wind power ultra-short-period prediction of a self- learning ARMA model in real time. The method includes the steps of inputting data to obtain auto regressive moving average model parameters, inputting wind resource monitoring system data and running monitoring system data, correcting startup capacity in real time according to the running monitoring data, establishing the auto regressive moving average model to obtain a wind power ultra-short-period prediction result, and introducing real-time anemometer tower data to correct the wind power ultra-short-period prediction result in real time. By predicting the wind power in the wind power generation process and introducing the real-time anemometer tower data to correct the wind power ultra-short-period prediction result in real time, the defect that in the existing ARMA technology, the wind power ultra-short-period prediction accuracy is low is overcome, and the aim of high-precision wind power ultra-short-period prediction is achieved.

Description

Ultra-short-term wind power forecasting method based on self-learning ARMA model with real-time correction

技术领域technical field

本发明涉及新能源发电过程中风电功率预测技术领域，具体地涉及一种测风网络实时校正的自学习ARMA模型风电功率超短期预测方法。The invention relates to the technical field of wind power forecasting in the process of new energy power generation, in particular to a self-learning ARMA model wind power ultra-short-term forecasting method for real-time correction of wind measuring networks.

背景技术Background technique

我国风电进入规模化发展阶段以后所产生的大型新能源基地多数位于“三北地区”(西北、东北、华北)，大型新能源基地一般远离负荷中心，其电力需要经过长距离、高电压输送到负荷中心进行消纳。由于风、光资源的间歇性、随机性和波动性，导致大规模新能源基地的风电、光伏发电出力会随之发生较大范围的波动，进一步导致输电网络充电功率的波动，给电网运行安全带来一系列问题。Most of the large-scale new energy bases generated after my country's wind power enters the stage of large-scale development are located in the "three north regions" (Northwest, Northeast, and North China). Large-scale new energy bases are generally far away from the load center, and their power needs to be transmitted to load center for consumption. Due to the intermittence, randomness and volatility of wind and light resources, the output of wind power and photovoltaic power generation in large-scale new energy bases will fluctuate in a large range, which will further lead to fluctuations in the charging power of the transmission network, which will affect the safety of power grid operation. bring a series of problems.

截至2014年4月，甘肃电网并网风电装机容量已达707万千瓦，约占甘肃电网总装机容量的22％，成为仅次于火电的第二大主力电源。目前，甘肃电网风电、光伏发电装机超过甘肃电网总装机容量的1/3。随着新能源并网规模的不断提高，风电、光伏发电不确定性和不可控性给电网的安全稳定经济运行带来诸多问题。准确预估可利用的发电风资源是对大规模风电优化调度的基础。对风力发电过程中的风电功率进行预测，可为新能源发电实时调度、新能源发电日前计划、新能源发电月度计划、新能源发电能力评估和弃风电量估计提供关键信息。As of April 2014, Gansu grid-connected wind power installed capacity has reached 7.07 million kilowatts, accounting for about 22% of the total installed capacity of Gansu grid, becoming the second largest main power source after thermal power. At present, the installed capacity of wind power and photovoltaic power generation in Gansu Power Grid exceeds 1/3 of the total installed capacity of Gansu Power Grid. With the continuous improvement of the grid-connected scale of new energy, the uncertainty and uncontrollability of wind power and photovoltaic power generation have brought many problems to the safe, stable and economical operation of the power grid. Accurate estimation of available wind resources for power generation is the basis for optimal scheduling of large-scale wind power. The prediction of wind power in the process of wind power generation can provide key information for real-time dispatch of new energy power generation, day-ahead planning of new energy power generation, monthly plan of new energy power generation, evaluation of new energy power generation capacity and estimation of abandoned wind power.

ARMA(自回归滑动平均模型)作为一种成熟的机器学习方法广泛应用于风电功率超短期预测。ARMA模型由自回归模型(AR)和滑动平均模型(MA)组成，采用对历史功率进行自回归运算及对白噪声序列进行滑动平均来预测未来0-4小时内的风电出力。ARMA方法有很多优点，因此广泛用于风电功率超短期预测，但ARMA最大的缺点就是其预测的滞后性，即当风电出力发生改变时，ARMA预测的结果的变化速度普遍慢于实际风电出力变化速度，因此，严重影响ARMA的预测精度。ARMA (Autoregressive Moving Average Model), as a mature machine learning method, is widely used in ultra-short-term forecasting of wind power. The ARMA model consists of an autoregressive model (AR) and a moving average model (MA). It uses autoregressive calculations on historical power and sliding averages on white noise sequences to predict wind power output within 0-4 hours in the future. The ARMA method has many advantages, so it is widely used in ultra-short-term forecasting of wind power, but the biggest disadvantage of ARMA is the hysteresis of its prediction, that is, when the wind power output changes, the change speed of the ARMA predicted result is generally slower than the actual wind power output change. Speed, therefore, strongly affects the predictive accuracy of ARMA.

发明内容Contents of the invention

本发明的目的在于，针对上述问题，提出一种实时校正的自学习ARMA模型风电功率超短期预测方法，以实现高精度风电功率超短期预测的优点。The purpose of the present invention is to address the above problems, to propose a real-time corrected self-learning ARMA model wind power ultra-short-term prediction method to achieve the advantages of high-precision wind power ultra-short-term prediction.

为实现上述目的，本发明采用的技术方案是：In order to achieve the above object, the technical scheme adopted in the present invention is:

一种实时校正的自学习ARMA模型风电功率超短期预测方法，包括输入数据得到自回归滑动平均模型参数；A real-time corrected self-learning ARMA model wind power ultra-short-term prediction method, including input data to obtain autoregressive moving average model parameters;

输入风资源监测系统数据和运行监测系统数据，并根据运行监测数据实时校正开机容量；Input wind resource monitoring system data and operation monitoring system data, and correct the starting capacity in real time according to the operation monitoring data;

建立自回归滑动平均模型从而得到风电功率超短期预测结果；Establish an autoregressive moving average model to obtain the ultra-short-term prediction results of wind power;

引入实时测风塔数据对风电功率超短期预测结果进行实时校正。Introduce real-time wind tower data to correct the ultra-short-term forecast results of wind power in real time.

根据本发明的优选实施例，所述输入数据得到自回归滑动平均模型参数包括，输入模型训练基础数据；According to a preferred embodiment of the present invention, said input data to obtain autoregressive moving average model parameters includes, input model training basic data;

模型定阶；model ordering;

采用矩估计方法对定阶的ARMA(p,q)模型参数进行估计。The parameters of fixed-order ARMA(p,q) model are estimated by moment estimation method.

根据本发明的优选实施例，所述输入模型训练基础数据，输入数据包括，历史风速数据和历史功率数据。According to a preferred embodiment of the present invention, the input model training basic data includes historical wind speed data and historical power data.

根据本发明的优选实施例，所述模型定阶具体为：According to a preferred embodiment of the present invention, the order determination of the model is specifically:

采用残差方差图法进行模型定阶，具体为设x_t为需要估计的项，x_t-1,x_t-2,...,x_t-n为已知历史功率序列，对于ARMA(p,q)模型，模型定阶即确定模型中参数p和q的值；The residual variance map method is used to determine the order of the model. Specifically, let x _t be the item to be estimated, x _t-1 , x _t-2 ,..., x _tn are known historical power sequences, and for ARMA(p, q) model, the order of the model is to determine the values of the parameters p and q in the model;

用系列阶数逐渐递增的模型拟合原始序列，每次都计算残差平方和然后画出阶数和的图形，当阶数由小增大时，会显著下降，达到真实阶数后的值会逐渐趋于平缓，甚至反而增大，Fits the original series with a model of increasing order of the series, computing the residual sum of squares each time Then plot the order and The graph of , when the order increases from small, will decrease significantly, and after reaching the true order The value will gradually become flat, or even increase instead,

＝拟合误差的平方和/(实际观测值个数-模型参数个数)， = sum of squares of fitting errors/(number of actual observations - number of model parameters),

实际观测值个数指拟合模型时实际使用的观察值项数，对于具有N个观察值的序列，拟合AR(p)模型，则实际使用的观察值最多为N-p，模型参数个数指所建立的模型中实际包含的参数个数，对于含有均值的模型，模型参数个数为模型阶数加1，对于N个观测值的序列，ARMA模型的残差估计式为：The number of actual observations refers to the number of observations actually used when fitting the model. For a sequence with N observations, when fitting the AR(p) model, the number of observations actually used is at most N-p, and the number of model parameters refers to The number of parameters actually included in the established model. For a model with a mean value, the number of model parameters is the model order plus 1. For a sequence of N observations, the residual estimation formula of the ARMA model is:

其中，Q为拟合误差的平方和函数，和θ_j(1≤j≤q)是模型系数，N是观测序列长度，是模型参数中的常数项。Among them, Q is the square sum function of the fitting error, and θ _j (1≤j≤q) are the model coefficients, N is the observation sequence length, is a constant term in the model parameters.

根据本发明的优选实施例，所述采用矩估计方法对定阶的ARMA(p,q)模型参数进行估计具体步骤为：According to a preferred embodiment of the present invention, the specific steps of estimating the fixed-order ARMA (p, q) model parameters using the moment estimation method are as follows:

将风电场历史功率数据利用数据序列x₁,x₂,...,x_t表示，其样本自协方差定义为The historical power data of the wind farm is represented by the data sequence x ₁ , x ₂ ,..., x _t , and its sample autocovariance is defined as

${\overset{^^}{γ γ}}_{k k} = = \frac{11}{n no} {Σ Σ}_{t t = = k k + + 11}^{n no} {x x}_{t t} {x x}_{t t - - k k},,$

其中，k＝0,1,2,...,n-1，x_t和x_t-k均为数据序列x₁,x₂,...,x_t中的数值；Wherein, k=0,1,2,...,n-1, x _t and x _tk are values in the data sequence x ₁ , x ₂ ,...,x _t ;

则 ${\hat{γ}}_{0} = \frac{1}{n} Σ_{t = 1}^{n} x_{t}^{2}$ but ${\hat{γ}}_{0} = \frac{1}{no} Σ_{t = 1}^{no} x_{t}^{2}$

则历史功率数据样本自相关函数为：Then the autocorrelation function of historical power data samples is:

${\overset{^^}{ρ ρ}}_{k k} = = \frac{{\overset{^^}{γ γ}}_{k k}}{{\overset{^^}{γ γ}}_{00}} = = \frac{\frac{11}{n no} {Σ Σ}_{t t = = k k + + 11}^{n no} {x x}_{t t} {x x}_{t t - - k k}}{\frac{11}{n no} {Σ Σ}_{t t = = 11}^{n no} {x x}_{t t}^{22}} = = \frac{{Σ Σ}_{t t = = k k + + 11}^{n no} {x x}_{t t} {x x}_{t t - - k k}}{{Σ Σ}_{t t = = 11}^{n no} {x x}_{t t}^{22}},,$

其中，k＝0,1,2,...,n-1；Among them, k=0,1,2,...,n-1;

AR部分的矩估计为，The moments of the AR part are estimated as,

令make

则协方差函数为Then the covariance function is

用的估计代替γ_k，use An estimate of γ instead of _k ,

可得参数 Available parameters

对MA(q)模型系数采用矩估计有For MA(q) model coefficients Estimated by moments

$γ_{0} (y_{t}) = (1 + θ_{1}^{2} + θ_{2}^{2} + . . . + θ_{q}^{2}) σ_{a}^{2}$ 直到 $γ_{0} ({the y}_{t}) = (1 + θ_{1}^{2} + θ_{2}^{2} + . . . + θ_{q}^{2}) σ_{a}^{2}$ until

${γ γ}_{k k} (({y the y}_{t t})) = = (({- - θ θ}_{k k} + + {θ θ}_{11} {θ θ}_{k k + + 11} + + . . . . . . + + {θ θ}_{q q - - k k} {θ θ}_{q q})) {σ σ}_{a a}^{22}$

其中k＝1,2,...,m，where k=1,2,...,m,

以上m+1个方程非线性方程，采用迭代法进行求解即得到自回归滑动平均模型参数。The above m+1 equations are nonlinear equations, and the iterative method is used to solve the autoregressive moving average model parameters.

根据本发明的优选实施例，According to a preferred embodiment of the present invention,

所述风资源监测系统数据包括与待预测风电场相关的测风塔所监测的实时测风数据及数值天气预报数据预测的风电场平均风速，所述运行监测系统数据是待预测风电场风机实时监测信息，包括风机实时停开机情况及机组浆距角状态信息。The data of the wind resource monitoring system includes the real-time wind measurement data monitored by the anemometer tower related to the wind farm to be predicted and the average wind speed of the wind farm predicted by the numerical weather forecast data. Monitoring information, including the real-time shutdown and startup status of the fan and the status information of the pitch angle of the unit.

根据本发明的优选实施例，还包括，According to a preferred embodiment of the present invention, it also includes,

将预测结果输出的步骤；The step of outputting the prediction result;

以及对预测结果后评估及模型修正的步骤。As well as the steps of post-evaluation of the prediction results and model correction.

根据本发明的优选实施例，所述自回归滑动平均模型为：According to a preferred embodiment of the present invention, the autoregressive moving average model is:

其中，和θ_j(1≤j≤q)是系数，α_t是白噪声序列。in, and θ _j (1≤j≤q) are coefficients, and α _t is a white noise sequence.

根据本发明的优选实施例，所述引入实时测风塔数据对风电功率超短期预测结果进行实时校正具体为：According to a preferred embodiment of the present invention, the introduction of real-time anemometer tower data to perform real-time correction on the ultra-short-term prediction results of wind power is specifically:

设t₁时刻，测风塔监测得到的风电场平均风速为v₁，数值天气预报数据预测的风电场平均风速为u₁，风电场的实际出力为p₁；下一个时间点t₂时刻，数值天气预报数据预测的风电场平均风速为u₂，则风电场平均风速v₂为，Suppose at time t ₁ , the average wind speed of the wind farm monitored by the anemometer tower is v ₁ , the average wind speed of the wind farm predicted by numerical weather forecast data is u ₁ , and the actual output of the wind farm is p ₁ ; at the next time point t ₂ , The average wind speed of the wind farm predicted by numerical weather forecast data is u ₂ , then the average wind speed v ₂ of the wind farm is,

v₂＝v₁+(u₂-u₁)v ₂ =v ₁ +(u ₂ -u ₁ )

则风电场功率预测的参数修正量为Then the parameter correction amount of wind farm power prediction is

$k = (\frac{v_{2} - v_{1}}{v_{1}} \times 100 %) \times p_{1}$ (v_t≠0时)。 $k = (\frac{v_{2} - v_{1}}{v_{1}} \times 100 %) \times p_{1}$ (when v _t ≠ 0).

根据本发明的优选实施例，输出的最终预测结果为：According to a preferred embodiment of the present invention, the final prediction result of the output is:

其中，X_t是t时刻风电场出力预测，和θ_j(1≤j≤q)是系数，α_t是白噪声序列，λ是加权系数，v_t是t时刻风电场的平均风速。Among them, X _t is the wind farm output forecast at time t, and θ _j (1≤j≤q) are coefficients, α _t is a white noise sequence, λ is a weighting coefficient, and v _t is the average wind speed of the wind farm at time t.

本发明的技术方案具有以下有益效果：The technical solution of the present invention has the following beneficial effects:

本发明的技术方案通过对风力发电过程中的风电功率进行预测，并通过引入实时测风塔数据对风电功率超短期预测结果进行实时校正，克服现有ARMA技术中风电功率超短期预测精度低的缺陷，达到高精度的风电功率超短期预测的目的。The technical solution of the present invention overcomes the defect of low ultra-short-term prediction accuracy of wind power in the existing ARMA technology by predicting the wind power in the process of wind power generation and correcting the ultra-short-term wind power prediction results in real time by introducing real-time anemometer data , to achieve the purpose of ultra-short-term forecasting of wind power with high precision.

下面通过附图和实施例，对本发明的技术方案做进一步的详细描述。The technical solutions of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

附图说明Description of drawings

图1为本发明实施例所述的实时校正的自学习ARMA模型风电功率超短期预测方法原理框图。Fig. 1 is a functional block diagram of a real-time corrected self-learning ARMA model wind power ultra-short-term prediction method according to an embodiment of the present invention.

具体实施方式Detailed ways

以下结合附图对本发明的优选实施例进行说明，应当理解，此处所描述的优选实施例仅用于说明和解释本发明，并不用于限定本发明。The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described here are only used to illustrate and explain the present invention, and are not intended to limit the present invention.

如图1所示，本发明技术方案提出的风电功率超短期预测可分为两个阶段：模型训练阶段和功率预测阶段。As shown in FIG. 1 , the ultra-short-term prediction of wind power proposed by the technical solution of the present invention can be divided into two stages: a model training stage and a power prediction stage.

阶段1：模型训练Phase 1: Model Training

步骤1.1：模型训练基础数据输入Step 1.1: Model training basic data input

风电功率预测系统模型训练所需输入数据主要包括历史风速数据、历史功率数据等。将基础数据输入到预测模型中进行模型训练。The input data required for model training of wind power forecasting system mainly include historical wind speed data, historical power data, etc. Input the basic data into the predictive model for model training.

步骤1.2：模型定阶Step 1.2: Model Ordering

由于事先无法确定需要使用多少已知时间序列的项来建立估计函数，所以需要对模型进行定阶判断。Since it is impossible to determine in advance how many items of known time series need to be used to establish the estimation function, it is necessary to make an order judgment on the model.

设x_t为需要估计的项，x_t-1,x_t-2,...,x_t-n为已知历史功率序列，对于ARMA(p,q)模型，模型定阶就是确定模型中参数p和q的值。Let x _t be the item to be estimated, x _t-1 , x _t-2 ,..., x _tn are the known historical power sequences, for the ARMA(p,q) model, model order determination is to determine the parameter p in the model and the value of q.

采用残差方差图法进行模型定阶。假定模型是有限阶自回归模型，如果设置的阶数小于真实阶数，则是一种不足拟合，因而拟合残差平方和必定偏大，此时通过提高阶数可以显著降低残差平方和。反之，如果阶数已经达到真实值，那么再增加阶数，就是过度拟合，此时增加阶数不会令残差平方和显著减小，甚至会略有增加。The order of the model was determined using the residual variogram method. Assuming that the model is a finite-order autoregressive model, if the set order is smaller than the true order, it is a kind of underfitting, so the sum of the squares of the fitting residuals must be too large. At this time, the residual squares can be significantly reduced by increasing the order and. Conversely, if the order has reached the true value, then increasing the order is overfitting. At this time, increasing the order will not significantly reduce the residual sum of squares, or even increase slightly.

这样用一系列阶数逐渐递增的模型来拟合原始序列，每次都计算残差平方和然后画出阶数和的图形。当阶数由小增大时，会显著下降，达到真实阶数后的值会逐渐趋于平缓，有时甚至反而增大。残差方差的估计式为：In this way, a series of models with increasing orders are used to fit the original sequence, and the sum of squared residuals is calculated each time Then plot the order and graphics. When the order increases from small to small, will decrease significantly, and after reaching the true order The value of will gradually level off, and sometimes even increase instead. The estimator of the residual variance is:

＝拟合误差的平方和/(实际观测值个数-模型参数个数) = sum of squares of fitting error/(number of actual observations - number of model parameters)

“实际观测值个数”是指拟合模型时实际使用的观察值项数，对于具有N个观察值的序列，拟合AR(p)模型，则实际使用的观察值最多为N-p。"Number of actual observations" refers to the number of observations actually used when fitting the model. For a sequence with N observations, when fitting the AR(p) model, the number of observations actually used is at most N-p.

“模型参数个数”是指所建立的模型中实际包含的参数个数，对于含有均值的模型，模型参数个数为模型阶数加1。对于N个观测值的序列，相应ARMA模型的残差估计式为："Number of model parameters" refers to the number of parameters actually included in the established model. For a model with a mean value, the number of model parameters is the model order plus 1. For a sequence of N observations, the residual estimation formula of the corresponding ARMA model is:

式1中，Q为拟合误差的平方和函数，和θ_j(1≤j≤q)是模型系数，N是观测序列长度，是模型参数中的常数项。In formula 1, Q is the sum of squares function of the fitting error, and θ _j (1≤j≤q) are the model coefficients, N is the observation sequence length, is a constant term in the model parameters.

步骤1.3：模型参数估计Step 1.3: Model parameter estimation

采用矩估计方法对ARMA(p,q)的模型参数进行估计。首先，将风电场历史功率数据利用数据序列x₁,x₂,...,x_t表示，其样本自协方差定义为The model parameters of ARMA(p,q) are estimated by moment estimation method. First, the historical power data of the wind farm is represented by the data sequence x ₁ , x ₂ ,..., x _t , and its sample autocovariance is defined as

${\hat{γ}}_{k} = \frac{1}{n} Σ_{t = k + 1}^{n} x_{t} x_{t - k}$ (式2) ${\hat{γ}}_{k} = \frac{1}{no} Σ_{t = k + 1}^{no} x_{t} x_{t - k}$ (Formula 2)

其中，k＝0,1,2,...,n-1，x_t和x_t-k均为数据序列x₁,x₂,...,x_t中的数值。Wherein, k=0,1,2,...,n-1, x _t and x _tk are values in the data sequence x ₁ , x ₂ ,...,x _t .

特别的，special,

${\hat{γ}}_{0} = \frac{1}{n} Σ_{t = 1}^{n} x_{t}^{2}$ (式3) ${\hat{γ}}_{0} = \frac{1}{no} Σ_{t = 1}^{no} x_{t}^{2}$ (Formula 3)

${\hat{ρ}}_{k} = \frac{{\hat{γ}}_{k}}{{\hat{γ}}_{0}} = \frac{\frac{1}{n} Σ_{t = k + 1}^{n} x_{t} x_{t - k}}{\frac{1}{n} Σ_{t = 1}^{n} x_{t}^{2}} = \frac{Σ_{t = k + 1}^{n} x_{t} x_{t - k}}{Σ_{t = 1}^{n} x_{t}^{2}}$ (式4) ${\hat{ρ}}_{k} = \frac{{\hat{γ}}_{k}}{{\hat{γ}}_{0}} = \frac{\frac{1}{no} Σ_{t = k + 1}^{no} x_{t} x_{t - k}}{\frac{1}{no} Σ_{t = 1}^{no} x_{t}^{2}} = \frac{Σ_{t = k + 1}^{no} x_{t} x_{t - k}}{Σ_{t = 1}^{no} x_{t}^{2}}$ (Formula 4)

其中，k＝0,1,2,...,n-1。Wherein, k=0, 1, 2, . . . , n-1.

AR部分的矩估计为The moments of the AR part are estimated as

(式5) (Formula 5)

令make

(式6) (Formula 6)

则协方差函数为Then the covariance function is

(式7) (Formula 7)

用的估计代替γ_k，有use The estimate of γ instead of _k , has

(式8) (Formula 8)

可得参数 Available parameters

对MA(q)模型系数θ₁,θ₂,...,θ_q采用矩估计有For MA(q) model coefficients θ ₁ , θ ₂ ,..., θ _q are estimated by moments

$γ_{0} (y_{t}) = (1 + θ_{1}^{2} + θ_{2}^{2} + . . . + θ_{q}^{2}) σ_{a}^{2}$ (式9) $γ_{0} ({the y}_{t}) = (1 + θ_{1}^{2} + θ_{2}^{2} + . . . + θ_{q}^{2}) σ_{a}^{2}$ (Formula 9)

………………

$γ_{k} (y_{t}) = ({- θ}_{k} + θ_{1} θ_{k + 1} + . . . + θ_{q - k} θ_{q}) σ_{a}^{2}$ (式10) $γ_{k} ({the y}_{t}) = ({- θ}_{k} + θ_{1} θ_{k + 1} + . . . + θ_{q - k} θ_{q}) σ_{a}^{2}$ (Formula 10)

其中k＝1,2,...,m。where k=1,2,...,m.

以上共包含m+1个方程，对其参数而言，方程为非线性，采用迭代法进行求解。The above contains a total of m+1 equations. For its parameters, the equations are nonlinear, and the iterative method is used to solve them.

具体步骤如下，将方程变形为：The specific steps are as follows, transforming the equation into:

$σ_{a}^{2} = γ_{0} / (1 + θ_{1}^{2} + θ_{2}^{2} + . . . + θ_{q}^{2})$ (式11) $σ_{a}^{2} = γ_{0} / (1 + θ_{1}^{2} + θ_{2}^{2} + . . . + θ_{q}^{2})$ (Formula 11)

$θ_{k} = - \frac{γ_{k}}{σ_{a}^{2}} + θ_{1} θ_{k + 1} + . . . + θ_{q - k} θ_{q}, k = 1,2, . . ., m$ (式12) $θ_{k} = - \frac{γ_{k}}{σ_{a}^{2}} + θ_{1} θ_{k + 1} + . . . + θ_{q - k} θ_{q}, k = 1,2, . . ., m$ (Formula 12)

给定θ₁,θ₂,...,θ_q和的一组初始值，如Given θ ₁ ,θ ₂ ,...,θ _q and A set of initial values, such as

$θ_{1} = θ_{2} = . . . = θ_{q} = 0, σ_{a}^{2} = γ_{0}$ (式13) $θ_{1} = θ_{2} = . . . = θ_{q} = 0, σ_{a}^{2} = γ_{0}$ (Formula 13)

代入以上两式右边，左边所得到的值为第一步迭代值，记为再将该值依次代入上两式的右侧，便得到第二步迭代值，依次类推，直到相邻两次迭代结果小于给定阈值时，取所得的结果作为参数的近似解。Substituting the right side of the above two equations, the value obtained on the left side is the value of the first iteration, denoted as Then substitute this value into the right side of the above two formulas in turn to get the second iteration value, By analogy, until the results of two adjacent iterations are less than the given threshold, the obtained results are taken as the approximate solution of the parameters.

通过上述求解过程发现，要求解时间序列模型的阶数，就要得到时间序列的预测值；要得到时间序列的预测值，必须先建立具体的预测函数；要建立具体的预测函数，必须知道模型的阶数。Through the above solution process, it is found that in order to solve the order of the time series model, the predicted value of the time series must be obtained; to obtain the predicted value of the time series, a specific prediction function must be established first; to establish a specific prediction function, the model must be known of order.

根据实践，时间序列模型阶数一般不超过5阶。所以在该算法具体实现时，可以首先假设模型为1阶，利用步骤1.3中的参数估计方法得到一阶模型的参数，进而建立估计函数便可以求得一阶模型时间序列模型估计得到各个项的预测值，从而求得一阶模型的残差方差；之后，假设模型为二阶，用上述方法求得二阶模型的残差；以此类推，可以得到1到5阶模型的残差，选残差最小的模型的阶数作为最终模型的阶数。确定模型阶数后，便可计算得到参数θ₁,θ₂,...,θ_q的值。According to practice, the order of the time series model generally does not exceed 5. Therefore, when implementing the algorithm, we can first assume that the model is first-order, use the parameter estimation method in step 1.3 to obtain the parameters of the first-order model, and then establish an estimation function to obtain the first-order model time series model and estimate the parameters of each item predicted value, so as to obtain the residual variance of the first-order model; then, assuming that the model is second-order, use the above method to obtain the residual error of the second-order model; The order of the model with the smallest residual is taken as the order of the final model. After the order of the model is determined, the values of the parameters θ ₁ , θ ₂ ,...,θ _q can be calculated.

阶段2：功率预测Phase 2: Power Prediction

步骤2.1：风资源监测系统数据输入Step 2.1: Wind resource monitoring system data input

风资源监测系统数据主要包括与待预测风电场相关的测风塔所监测的实时测风数据及NWP(数值天气预报数据)预测的风电场平均风速。The data of the wind resource monitoring system mainly include the real-time wind measurement data monitored by the anemometer towers related to the wind farm to be predicted and the average wind speed of the wind farm predicted by NWP (numerical weather prediction data).

步骤2.2：运行监测系统数据输入Step 2.2: Operation Monitoring System Data Entry

运行监测系统数据是指待预测风电场风机实时监测信息，主要包括风机实时停开机情况及机组浆距角等状态信息。The data of the operation monitoring system refers to the real-time monitoring information of the wind turbines of the wind farm to be predicted, mainly including the real-time shutdown and start-up of the wind turbines and the status information such as the pitch angle of the units.

步骤2.3：运行监测数据实时校正开机容量Step 2.3: Real-time correction of power-on capacity based on operation monitoring data

由于风电场运行过程中，总有各种原因所导致的停机情况，比如典型的20万千瓦装机的风电场，包含134台风机，平均有10台左右的风机处于停机状态，因此通过实时风机运行监测数据可以知道风电实际的开机容量，而非用风电场的装机容量进行风电功率超短期预测。During the operation of the wind farm, there are always shutdowns caused by various reasons. For example, a typical wind farm with 200,000 kilowatts installed capacity contains 134 wind turbines, and an average of about 10 wind turbines are in shutdown state. Therefore, real-time wind turbine operation The monitoring data can know the actual start-up capacity of wind power, instead of using the installed capacity of wind farms for ultra-short-term forecasting of wind power.

步骤2.4：基于ARMA模型的风电功率超短期预测Step 2.4: Ultra-short-term forecast of wind power based on ARMA model

将模型参数估计出来之后，结合已估计的模型阶数，便可得到用于风电功率超短期预测的时间序列方程。根据上述步骤2和步骤3得出的p和q值，以及θ₁,θ₂,...,θ_q的值建立自回归滑动平均模型；After estimating the model parameters, combined with the estimated model order, the time series equations for ultra-short-term forecasting of wind power can be obtained. p and q values derived from steps 2 and 3 above, and The value of θ ₁ , θ ₂ ,..., θ _q establishes an autoregressive moving average model;

自回归滑动平均模型如下：The autoregressive moving average model is as follows:

(式14) (Formula 14)

其中，和θ_j(1≤j≤q)是系数，αt是白噪声序列。in, and θ _j (1≤j≤q) are coefficients, and αt is a white noise sequence.

步骤2.5：资源监测数据实时校正风电功率超短期预测结果Step 2.5: Real-time correction of wind power ultra-short-term forecast results based on resource monitoring data

通过上述ARMA预测模型可以看出，对于风电功率的实时变化，上述模型总是具有滞后性，本发明通过引入实时测风塔数据对风电功率超短期预测结果进行实时校正。It can be seen from the above ARMA prediction model that the above-mentioned model always has hysteresis for real-time changes in wind power. The present invention corrects the ultra-short-term prediction results of wind power in real time by introducing real-time anemometer data.

设t₁时刻，测风塔监测得到的风电场平均风速为v₁，NWP预测的风电场平均风速为u₁，风电场的实际出力为p₁；下一个时间点t₂时刻，NWP预测的风电场平均风速为u₂，则风电场平均风速v₂为，Suppose at time t ₁ , the average wind speed of the wind farm monitored by the anemometer tower is v ₁ , the average wind speed of the wind farm predicted by NWP is u ₁ , and the actual output of the wind farm is p ₁ ; at the next time point t ₂ , the NWP predicted The average wind speed of the wind farm is u ₂ , then the average wind speed v ₂ of the wind farm is,

v₂＝v₁+(u₂-u₁) (式15)v ₂ =v ₁ +(u ₂ -u ₁ ) (Equation 15)

$k = (\frac{v_{2} - v_{1}}{v_{1}} \times 100 %) \times p_{1}$ (v_t≠0时) (式16) $k = (\frac{v_{2} - v_{1}}{v_{1}} \times 100 %) \times p_{1}$ (when v _t ≠0) (Equation 16)

步骤2.6：最终预测结果输出及展示Step 2.6: Final forecast output and display

则测风网络实时校正的ARMA模型风电功率超短期预测结果为Then the ultra-short-term wind power prediction result of the ARMA model corrected by the wind measurement network in real time is

(式17) (Formula 17)

通过引入测风塔实时监测数据修正后的预测风速，可以对ARMA模型下一步预测做出加权调整，从而解决ARMA模型预测的滞后性问题。By introducing the corrected forecast wind speed from the real-time monitoring data of the wind measuring tower, weighted adjustments can be made to the next forecast of the ARMA model, thereby solving the lagging problem of the ARMA model forecast.

将预测结果输出至数据库中，并通过图表及曲线展示预测结果、展示预测与实测结果的对比。Output the forecast results to the database, and display the forecast results through charts and curves, and show the comparison between the forecast and actual measurement results.

步骤2.7：预测结果后评估及模型修正Step 2.7: Post-evaluation of prediction results and model correction

该步骤首先对预测结果进行后评估，分析预测值与实测值之间的误差。如果预测误差大于允许的最大误差，则跳转到模型训练过程，重新进行模型训练。In this step, post-evaluation is first performed on the prediction results, and the error between the predicted value and the measured value is analyzed. If the prediction error is greater than the maximum allowable error, jump to the model training process and perform model training again.

最后应说明的是：以上所述仅为本发明的优选实施例而已，并不用于限制本发明，尽管参照前述实施例对本发明进行了详细的说明，对于本领域的技术人员来说，其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换。凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。Finally, it should be noted that: the above is only a preferred embodiment of the present invention, and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, it still The technical solutions recorded in the foregoing embodiments may be modified, or some technical features thereof may be equivalently replaced. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A real-time correction ultra-short-term wind power prediction method for a self-learning ARMA model is characterized by comprising the steps of inputting data to obtain parameters of an autoregressive moving average model;

inputting wind resource monitoring system data and operation monitoring system data, and correcting the starting capacity in real time according to the operation monitoring data;

establishing an autoregressive moving average model so as to obtain a wind power ultra-short term prediction result;

and introducing real-time anemometer tower data to correct the wind power ultra-short term prediction result in real time.

2. The real-time corrected self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 1, wherein the obtaining of autoregressive moving average model parameters from the input data comprises inputting model training base data;

determining the order of the model;

and (3) estimating the fixed-order ARMA (p, q) model parameters by adopting a moment estimation method.

3. The real-time corrected self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 2, wherein the input model trains basic data, and the input data comprises historical wind speed data and historical power data.

4. The real-time correction self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 3, wherein the model order determination specifically comprises:

performing model order determination by using a residual variance graph method, specifically setting x_tFor the term to be estimated, x_t-1,x_t-2,...,x_t-nFor an ARMA (p, q) model, determining the values of parameters p and q in the model by the model in order for the known historical power sequence;

fitting the original sequence with a model with a series of increasing orders, calculating the sum of squares of the residuals each timeThen draw the sum of the ordersWhen the order number is increased from small to small,will be obviously reduced and reach the real orderThe value of (a) will gradually become flat, or even increase,

the square sum of the fitting errors/(number of actual observed values-number of model parameters),

the number of actual observed values refers to the number of observed value terms actually used in fitting the model, for a sequence with N observed values, fitting an AR (p) model, the actually used observed values are at most N-p, the model parameter number refers to the number of parameters actually contained in the established model, for the model with a mean value, the number of model parameters is the number of model orders plus 1, and for the sequence with N observed values, the residual estimation formula of the ARMA model is as follows:

wherein Q is a sum of squares function of the fitting error,and theta_j(1. ltoreq. j. ltoreq. q) is the model coefficient, N is the observation sequence length,is a constant term in the model parameters.

5. The real-time correction self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 4, wherein the specific steps of estimating fixed-order ARMA (p, q) model parameters by using a moment estimation method are as follows:

utilizing historical power data of wind power plant by data sequence x₁,x₂,...,x_tRepresentation with sample autocovariance defined as

Wherein k is 0,1,2_tAnd x_t-kAre all data sequences x₁,x₂,...,x_tThe numerical values of (1);

then

The historical power data sample autocorrelation function is then:

wherein k is 0,1,2, 1, n-1;

the moment of the AR part is estimated as,

order to

The covariance function is then

By usingInstead of gamma_k，

Available parameters

For the MA (q) model coefficient theta₁,θ₂,...,θ_qUsing the moment estimate to have

Up to

Wherein k is 1, 2.. times, m,

and solving the nonlinear equations of the above m +1 equations by an iterative method to obtain the parameters of the autoregressive moving average model.

6. The real-time corrected self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 5,

the wind resource monitoring system data comprises real-time wind measurement data monitored by a wind measurement tower related to the wind power plant to be predicted and the average wind speed of the wind power plant predicted by numerical weather forecast data, and the operation monitoring system data is real-time monitoring information of a fan of the wind power plant to be predicted and comprises real-time startup and shutdown conditions of the fan and unit pitch angle state information.

7. The real-time corrected self-learning ARMA model wind power ultra-short term prediction method as recited in claim 6, further comprising,

outputting the prediction result;

and post-evaluating the prediction result and modifying the model.

8. The real-time corrected self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 7, wherein the autoregressive moving average model is:

wherein,and theta_j(1. ltoreq. j. ltoreq. q) is a coefficient, alpha_tIs a white noise sequence.

9. The real-time correction self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 8, wherein the real-time correction of the wind power ultra-short term prediction result by introducing real-time anemometer tower data specifically comprises:

let t₁At any moment, the average wind speed of the wind power plant obtained by monitoring the anemometer tower is v₁And the average wind speed of the wind power plant predicted by numerical weather forecast data is u₁The actual output of the wind farm is p₁(ii) a The next time t₂At the moment, the average wind speed of the wind power plant predicted by numerical weather forecast data is u₂Then the average wind speed v of the wind farm₂In order to realize the purpose,

v₂＝v₁+(u₂-u₁)

the correction quantity of the parameter of the wind power plant power prediction is

(v_tNot equal to 0).

10. The real-time correction self-learning ARMA model wind power ultra-short term prediction method as claimed in claim 9, wherein the final output prediction result is:

wherein, X_tIs the prediction of the output of the wind power plant at the moment t,and theta_j(1. ltoreq. j. ltoreq. q) is a coefficient, alpha_tIs a white noise sequence, λ is a weighting coefficient, v_tIs the average wind speed of the wind farm at time t.