Spatio‐temporal analysis and modeling of short‐term wind power forecast errors

Forecasts of wind power production are increasingly being used in various management tasks. So far, such forecasts and related uncertainty information have usually been generated individually for a given site of interest (either a wind farm or a group of wind farms), without properly accounting for the spatio‐temporal dependencies observed in the wind generation field. However, it is intuitively expected that, owing to the inertia of meteorological forecasting systems, a forecast error made at a given point in space and time will be related to forecast errors at other points in space in the following period. The existence of such underlying correlation patterns is demonstrated and analyzed in this paper, considering the case‐study of western Denmark. The effects of prevailing wind speed and direction on autocorrelation and cross‐correlation patterns are thoroughly described. For a flat terrain region of small size like western Denmark, significant correlation between the various zones is observed for time delays up to 5 h. Wind direction is shown to play a crucial role, while the effect of wind speed is more complex. Nonlinear models permitting capture of the interdependence structure of wind power forecast errors are proposed, and their ability to mimic this structure is discussed. The best performing model is shown to explain 54% of the variations of the forecast errors observed for the individual forecasts used today. Even though focus is on 1‐h‐ahead forecast errors and on western Denmark only, the methodology proposed may be similarly tested on the cases of further look‐ahead times, larger areas, or more complex topographies. Such generalization may not be straightforward. While the results presented here comprise a first step only, the revealed error propagation principles may be seen as a basis for future related work. Copyright © 2010 John Wiley & Sons, Ltd.     

Comparative analysis of methods for modelling the short‐term probability distribution of extreme wind turbine loads

We have tested the performance of statistical extrapolation methods in predicting the extreme response of a multi‐megawatt wind turbine generator. We have applied the peaks‐over‐threshold, block maxima and average conditional exceedance rates (ACER) methods for peaks extraction, combined with four extrapolation techniques: the Weibull, Gumbel and Pareto distributions and a double‐exponential asymptotic extreme value function based on the ACER method. For the successful implementation of a fully automated extrapolation process, we have developed a procedure for automatic identification of tail threshold levels, based on the assumption that the response tail is asymptotically Gumbel distributed. Example analyses were carried out, aimed at comparing the different methods, analysing the statistical uncertainties and identifying the factors, which are critical to the accuracy and reliability of the extrapolation. The present paper describes the modelling procedures and makes a comparison of extrapolation methods based on the results from the example calculations. Copyright © 2015 John Wiley & Sons, Ltd.     

Implementation of a Model Output Statistics based on meteorological variable screening for short‐term wind power forecast

A combination of physical and statistical treatments to post‐process numerical weather predictions (NWP) outputs is needed for successful short‐term wind power forecasts. One of the most promising and effective approaches for statistical treatment is the Model Output Statistics (MOS) technique. In this study, a MOS based on multiple linear regression is proposed: the model screens the most relevant NWP forecast variables and selects the best predictors to fit a regression equation that minimizes the forecast errors, utilizing wind farm power output measurements as input. The performance of the method is evaluated in two wind farms, located in different topographical areas and with different NWP grid spacing. Because of the high seasonal variability of NWP forecasts, it was considered appropriate to implement monthly stratified MOS. In both wind farms, the first predictors were always wind speeds (at different heights) or friction velocity. When friction velocity is the first predictor, the proposed MOS forecasts resulted to be highly dependent on the friction velocity–wind speed correlation. Negligible improvements were encountered when including more than two predictors in the regression equation. The proposed MOS performed well in both wind farms, and its forecasts compare positively with an actual operative model in use at Risø DTU and other MOS types, showing minimum BIAS and improving NWP power forecast of around 15% in terms of root mean square error. Further improvements could be obtained by the implementation of a more refined MOS stratification, e.g. fitting specific equations in different synoptic situations. Copyright © 2012 John Wiley & Sons, Ltd.     

基于自适应阈值的小波BP神经网络风电变桨系统故障预测

为了提高风能资源的有效利用,提高风电机组运行的可靠性、经济性和安全性,故障预测变得尤为重要。故障预测方法在判断设备隐患、制定合理的风电场运维方案方面具有重要的理论和实际意义。围绕变桨系统故障预测的问题,文章利用小波对机械信号特征敏感的优点,引入自适应阈值函数实现对小波降噪的改进,结合具有自学习能力和并行处理能力强的BP神经网络,建立了自适应阈值的小波BP神经网络故障预测模型。该模型结合了小波分析的技术特点,减少了噪声对预测模型的干扰,模型简洁、易实现。应用该网络预测模型,提前15 d对变桨系统故障预测的准确率达到了92.27%。     

LASSO vector autoregression structures for very short‐term wind power forecasting

The deployment of smart grids and renewable energy dispatch centers motivates the development of forecasting techniques that take advantage of near real‐time measurements collected from geographically distributed sensors. This paper describes a forecasting methodology that explores a set of different sparse structures for the vector autoregression (VAR) model using the least absolute shrinkage and selection operator (LASSO) framework. The alternating direction method of multipliers is applied to fit the different LASSO‐VAR variants and create a scalable forecasting method supported by parallel computing and fast convergence, which can be used by system operators and renewable power plant operators. A test case with 66 wind power plants is used to show the improvement in forecasting skill from exploring distributed sparse structures. The proposed solution outperformed the conventional autoregressive and vector autoregressive models, as well as a sparse VAR model from the state of the art. Copyright © 2016 John Wiley & Sons, Ltd.     

From probabilistic forecasts to statistical scenarios of short‐term wind power production

Short‐term (up to 2–3 days ahead) probabilistic forecasts of wind power provide forecast users with highly valuable information on the uncertainty of expected wind generation. Whatever the type of these probabilistic forecasts, they are produced on a per horizon basis, and hence do not inform on the development of the forecast uncertainty through forecast series. However, this additional information may be paramount for a large class of time‐dependent and multistage decision‐making problems, e.g. optimal operation of combined wind‐storage systems or multiple‐market trading with different gate closures. This issue is addressed here by describing a method that permits the generation of statistical scenarios of short‐term wind generation that accounts for both the interdependence structure of prediction errors and the predictive distributions of wind power production. The method is based on the conversion of series of prediction errors to a multivariate Gaussian random variable, the interdependence structure of which can then be summarized by a unique covariance matrix. Such matrix is recursively estimated in order to accommodate long‐term variations in the prediction error characteristics. The quality and interest of the methodology are demonstrated with an application to the test case of a multi‐MW wind farm over a period of more than 2 years. Copyright © 2008 John Wiley & Sons, Ltd.     

基于CEEMD-CNN-BiGRU-RF模型的短期风电功率预测

风电功率的准确预测对电网的安全运行和经济调度起着重要作用,为进一步提高风电功率的预测精度,文章提出了一种基于CEEMD-CNN-BiGRU-RF模型的短期风电功率预测模型。首先,利用完全集成经验模态分解(CEEMD)对风电功率时间序列进行模态分解;其次,对分解的各个风电功率时间序列利用卷积神经网络(CNN)进行特征提取;再次,建立双向门控循环单元(Bi GRU)模型对各个风电功率时间序列进行预测,叠加各个分量的预测值;最后,对误差进行进一步分析与预测,利用随机森林(RF)进行误差修正,得到最终的风电功率预测值。实验仿真表明,该模型的预测效果明显优于传统模型,模型的平均绝对百分比误差(MAPE)仅为2.09%。     

A prediction approach using ensemble empirical mode decomposition‐permutation entropy and regularized extreme learning machine for short‐term wind speed

Accurate prediction of short‐term wind speed is of great significance to the operation and maintenance of wind farms, the optimal scheduling of turbines, and the safe and stable operation of power grids. A prediction approach for short‐term wind speed using ensemble empirical mode decomposition‐permutation entropy and regularized extreme learning machine is proposed. Firstly, wind speed time series is decomposed into several components with different frequency by ensemble empirical mode decomposition, which can reduce the non‐stationarity of the original time series. The permutation entropy value for each component is used to analyze its complexity. The components can be recombined to obtain a set of new subsequences. Then, different prediction models based on regularized extreme learning machine are used to predict each subsequence. Fivefold cross validation is used to improve the reliability of the regularized extreme learning machine model. Finally, the predicted value of each subsequence is superimposed to obtain the final predictive result. Ten minutes, 30 minutes, and 1 hour short‐term wind speed data from wind farms in Liaoning Province, China, are used for conducting experiments. The experimental results indicate that the values of the root mean square error of the developed prediction approach utilizing 10 minutes, 30 minutes, and 1 hour interval data are 0.5629, 0.4473, and 0.5697; mean absolute error are 0.4427, 3.0701, and 0.4897; mean absolute percentile error are 4.1456%, 16.8166%, and 6.8166%; relative root mean square are 0.0505, 0.2997, and 0.2609; square sum error are 55.5263, 59.6347, and 64.9154; and the Theil inequality coefficient are 0.0235, 0.0808, and 0.0625, which are much lower than those of the comparison methods. The values of the R square of the developed prediction approach utilizing 10 minutes, 30 minutes, and 1 hour interval data are 0.9363, 0.9161, and 0.9472, and the index of agreement are 0.9994, 0.9925, and 0.9894, which are higher than those of the comparison methods. The Pearson's test results show that the association strength between the actual value and the predicted values of the proposed approach is stronger. Also, the proposed prediction approach in this paper has higher reliability under the same confidence level. The effectiveness of the proposed prediction approach for short‐term wind speed is verified.     

风电场微观选址及发电量预测的探讨

对风资源评估、选址地面情况和风机位置的排布等影响风电场微观选址的因素进行了分析.阐述了风电场发电量的预测方法,通过实例说明如何使用相关软件来预测风电场发电量,并根据预测结果对风电场微观选址注意事项进行了探讨.     

Generating short‐term probabilistic wind power scenarios via nonparametric forecast error density estimators

Forecasts of available wind power are critical in key electric power systems operations planning problems, including economic dispatch and unit commitment. Such forecasts are necessarily uncertain, limiting the reliability and cost‐effectiveness of operations planning models based on a single deterministic or “point” forecast. A common approach to address this limitation involves the use of a number of probabilistic scenarios, each specifying a possible trajectory of wind power production, with associated probability. We present and analyze a novel method for generating probabilistic wind power scenarios, leveraging available historical information in the form of forecasted and corresponding observed wind power time series. We estimate nonparametric forecast error densities, specifically using epi‐spline basis functions, allowing us to capture the skewed and nonparametric nature of error densities observed in real‐world data. We then describe a method to generate probabilistic scenarios from these basis functions that allows users to control for the degree to which extreme errors are captured. We compare the performance of our approach to the current state‐of‐the‐art considering publicly available data associated with the Bonneville Power Administration, analyzing aggregate production of a number of wind farms over a large geographic region. Finally, we discuss the advantages of our approach in the context of specific power systems operations planning problems: stochastic unit commitment and economic dispatch. Our methodology is embodied in the joint Sandia–University of California Davis Prescient software package for assessing and analyzing stochastic operations strategies.     

Risk‐based optimal distribution of power reserves in wind power plants

This paper employs the Conditional Value‐at Risk, largely used in financial risk management, to specify the power reserve capacity of a wind power plant (WPP) under a risk metric. Evidences are shown here that other popular, simpler measure, the Value‐at Risk, is inappropriate for that specification. Under this risk‐based reserve metric, two programs are approached to optimally distribute a reserve request in a WPP subject to a given confidence level in the commitment. The most exhaustive of the two is a two‐level formulation including a solution to the load power flow (LPF) in the WPP. By solving these two programs, for comparison with interior‐point and heuristic solvers, conclusions are drawn. Notably, that a Pareto optimality occurs for stringent reserve requests; that putting off‐line generators is financially more profitable than partial curtailments to respond to low reserve requests; and that in these cases accounting for losses through LPF‐based optimization seems unnecessary. Copyright © 2016 John Wiley & Sons, Ltd.     

Assessing the suitability of wind speed probabilty distribution functions based on wind power density

Three functions have so far predominantly been used for fitting the measured wind speed probability distribution in a given location over a certain period of time, typically monthly or yearly. In the literature, it is common to fit these functions to compare which one fits the measured distribution best in a particular location. During this comparison process, parameters on which the suitability of the fit is judged are required. The parameters that are mostly used are the mean wind speed or the total wind energy output (primary parameters). It is, however, shown in the present study that one cannot judge the suitability of the functions based on the primary parameters alone. Additional parameters (secondary parameters) that complete the primary parameters are required to have a complete assessment of the fit, such as the discrepancy between the measured and fitted distributions, both for the wind speed and wind energy (that is the standard deviation of wind speed and wind energy distributions). Therefore, the secondary statistical parameters have to be known as well as the primary ones to make a judgement about the suitability of the distribution functions analysed. The primary and secondary parameters are calculated from the 12-month of measured hourly wind speed data and detailed analyses of wind speed distributions are undertaken in the present article.     

Short‐term spatio‐temporal prediction of wind speed and direction

This paper aims to produce a low‐complexity predictor for the hourly mean wind speed and direction from 1 to 6 h ahead at multiple sites distributed around the UK. The wind speed and direction are modelled via the magnitude and phase of a complex‐valued time series. A multichannel adaptive filter is set to predict this signal on the basis of its past values and the spatio‐temporal correlation between wind signals measured at numerous geographical locations. The filter coefficients are determined by minimizing the mean square prediction error. To account for the time‐varying nature of the wind data and the underlying system, we propose a cyclo‐stationary Wiener solution, which is shown to produce an accurate predictor. An iterative solution, which provides lower computational complexity, increased robustness towards ill‐conditioning of the data covariance matrices and the ability to track time‐variations in the underlying system, is also presented. The approaches are tested on wind speed and direction data measured at various sites across the UK. Results show that the proposed techniques are able to predict wind speed as accurately as state‐of‐the‐art wind speed forecasting benchmarks while simultaneously providing valuable directional information. Copyright © 2013 John Wiley & Sons, Ltd.     

基于风电机组无功裕度预测的风电场无功分层控制策略

针对风电电压波动的问题,文章基于风电机组无功裕度预测,提出了一种风电场无功分层控制策略.该策略首先以并网点电压偏差和线路有功损耗最小为目标,使用二次规划算法在线实时求解最优并网电压,进而求解风电场无功参考值;其次,采用EWT-LSSVM预测算法进行风电功率预测,并提出预测功率校正方法实时修正预测功率,精确求解风电机组的...     

Trading wind power through physically settled options and short‐term electricity markets

Wind power producers participating in today's electricity markets face significant variability in revenue streams, with potential high losses mostly due to wind's limited predictability and the intermittent nature of the generated electricity. In order to further expand wind power generation despite such challenges, it is important to maximize its market value and move decisively towards economically sustainable and financially viable asset management. In this paper, we introduce a decision‐making framework based on stochastic optimization that allows wind power producers to hedge their position in the market by trading physically settled options in futures markets in conjunction with their participation in the short‐term electricity markets. The proposed framework relies on a series of two‐stage stochastic optimization models that identify a combined trading strategy for wind power producers actively participating in both financial and day‐ahead electricity markets. The proposed models take into consideration penalties from potential deviations between day‐ahead market offers and real‐time operation and incorporates different preferences of risk aversion, enabling a trade‐off between the expected profit and its variability. Empirical analysis based on data from the Nordic region illustrates high efficiency of the stochastic model and reveals increased revenues for both risk neutral and risk averse wind producers opting for combined strategies.     

基于BP神经网络的小型风力发电正弦波逆变器

为了提高小型风力发电系统的可靠性和能量转换效率,文章设计了一种带有高频环节的单相正弦逆变器,该逆变器提出采用双BP神经网络控制。在Matlab下建立了逆变器仿真模型,仿真结果表明,设计的BP神经网络控制器可以使单相正弦逆变器具有较高的稳态精度和动态特性,满足小型风力发电系统的需要。     

LiDAR‐based characterization of mid‐altitude wind conditions for airborne wind energy systems

Engineers and researchers working on the development of airborne wind energy systems (AWES) still rely on oversimplified wind speed approximations and coarsely sampled reanalysis data because of a lack of high‐resolution wind data at altitudes above 200 m. Ten‐minute average wind speed LiDAR measurements up to an altitude of 1100 m and data from nearby weather stations were investigated with regard to wind energy generation and impact on LiDAR measurements. Data were gathered by a long‐range pulsed Doppler LiDAR device installed on flat terrain. Because of the low overall carrier‐to‐noise ratio, a custom‐filtering technique was applied. Our analyses show that diurnal variation and atmospheric stability significantly affect wind conditions aloft which cause a wide range of wind speeds and a multimodal probability distribution that cannot be represented by a simple Weibull distribution fit. A better representation of the actual wind conditions can be achieved by fitting Weibull distributions separately to stable and unstable conditions. Splitting and clustering the data by simulated surface heat flux reveals substate stratification responsible for the multimodality. We classify different wind conditions based on these substates, which result in different wind energy potential. We assess optimal traction power and optimal operating altitudes statistically as well as for specific days based on a simplified AWES model. Using measured wind speed standard deviation, we estimate average turbulence intensity and show its variation with altitude and time. Selected short‐term data sets illustrate temporal changes in wind conditions and atmospheric stratification with a high temporal and vertical resolution.     

Forecasting method of monthly wind power generation based on climate model and long short-term memory neural network

Predicting wind power generation over the medium and long term is helpful for dispatching departments, as it aids in constructing generation plans and electricity market transactions. This study presents a monthly wind power generation forecasting method based on a climate model and long short-term memory (LSTM) neural network. A nonlinear mapping model is established between the meteorological elements and wind power monthly utilization hours. After considering the meteorological data (as predicted for the future) and new installed capacity planning, the monthly wind power generation forecast results are output. A case study shows the effectiveness of the prediction method.     

A novel method based on Weibull distribution for short-term wind speed prediction

Wind speed prediction (WSP) is essential in order to predict and analyze efficiency and performance of wind-based electricity generation systems. More accurate WSP may provide better opportunities to design and build more efficient and robust wind energy systems. Precious short-term prediction is difficult to achieve; therefore several methods have been developed so far. We notice that the statistics of the alterations, which occur between sequential values of the predicted wind speed data, may differ significantly from observed wind statistics. In this study, we investigate these alterations and compare them and, accordingly, propose a novel method based on Weibull and Gaussian probability distribution functions (PDF) for short-term WSP. The proposed method stands on an algorithm, which examines comparison of the statistical features of the observed and generated wind speed in order to achieve more accurate estimation. We have examined this method on the wind speed data set observed and recorded in Ankara in 2013 and in 2014. The obtained results show that the new algorithm provides better wind speed prediction with an enhanced wind speed model.     

基于时空图卷积神经网络的光伏发电功率超短期预测方法

为了解决传统光伏电站超短期功率预测方法不能同时准确提取发电功率的时间和空间特征的问题,提出一种基于时空图卷积神经网络的光伏发电功率超短期预测方法。针对同一区域内的多个光伏电站,首先对电站进行图建模,利用图卷积网络（GCN）与门控线性单元（GLU）提取发电功率的时空特征。利用提取到的时空特征信息以及区域内光伏电站的历史发电功率数据训练预测模型,最终实现对多个光伏电站发电功率超短期预测。实验结果表明,该方法能够将超短期功率预测均方根误差减小至1.122%,对工作人员根据实际情况进行电网的调度管理具有重要意义。