实数编码的演化算法求解TSP问题

对新近提出的求解TSP问题的实数编码的染色体表示方式进行了研究，为了去除存在于这种染色体表示方式中的冗余，对其进行了改动，然后设计了相应的多父体杂交算子和变异算子，完成了一个实数编码的求解TSP问题的演化算法。实验结果表明，这个算法是可行的，能够使解收敛到一定的程度，但还需要提高其收敛的能力。所以下一步的工作重点在于根据这种染色体表示方式的特点，进一步研究更合适的算子，从而得到更好的解。     

A new wind retrieval algorithm for Jason-1 at high wind speeds

The altimeter wind speed algorithm at high wind speeds remains unsolved because of lack of observed data. In this study data at high wind speeds were generated using Yin's typhoon model, which consists of the Rankine vortex model and the angular momentum model with typhoon parameters, provided by the Joint Typhoon Warning Centre (JTWC). The accuracy of Yin's typhoon model can be validated by comparing it with recorded data from a weather station. By comparing the normalized radar backscatter cross‐section (NRCS) detected by the Jason‐1 altimeter with wind speed data inferred by Yin's typhoon model, an empirical algorithm valid for a range of wind speeds between 10 and 40 m s^–1 is developed and proposed. The proposed algorithm is compared with the Jason‐1 operational algorithm and Young's altimeter wind retrieval algorithm. The study shows that, for the proposed algorithm and the operational algorithm for Jason‐1, the root mean square (RMS) errors are 3.38 and 3.60 m s^–1, respectively, and the average relative errors are 18% and 19%, respectively, for wind speeds less than 27 m s^–1. Hence, the proposed algorithm is in agreement with the operational algorithm for the Jason‐1 altimeter for wind speeds in the range 10–27 m s^–1. However, the Jason‐1 operational algorithm is inaccurate for wind speeds above 27 m s^–1 because the wind speeds used in the algorithm training process came from scatterometer wind products, and are significantly lower than those in strong wind and heavy rain conditions. Comparison of the proposed algorithm and Young's algorithm shows that the RMS errors are 6.27 and 15.18 m s^–1, respectively, and the average relative errors are 16% and 59%, respectively, for wind speeds greater than 20 m s^–1. The Holland typhoon model cannot accurately determine the outer wind field of typhoons since it extends cyclonic wind speeds to infinity. Young's altimeter wind retrieval algorithm depends on the Holland typhoon model, and the latter results in some errors. Compared with Young's altimeter wind retrieval algorithm, the proposed algorithm retrieves wind speeds with better accuracy. Therefore, the proposed algorithm, suitable for retrieving sea surface wind speeds in typhoons and other strong wind conditions, can be considered as supplementary to the Jason‐1 operational algorithm.     

Reliability-based design optimization of wind turbine drivetrain with integrated multibody gear dynamics simulation considering wind load uncertainty

This study aims to develop an integrated computational framework for the reliability-based design optimization (RBDO) of wind turbine drivetrains to assure the target reliability under wind load and gear manufacturing uncertainties. Gears in wind turbine drivetrains are subjected to severe cyclic loading due to highly variable wind loads that are stochastic in nature. Thus, the failure rate of drivetrain systems is reported to be higher than the other wind turbine components, and improving drivetrain reliability is critically important in reducing downtime caused by gear failures. In the numerical procedure developed in this study, a wide spatiotemporal variability for wind loads is considered using 249 sets of wind data to evaluate probabilistic contact fatigue life in the sampling-based RBDO. To account for wind load uncertainty in evaluation of the tooth contact fatigue, multiple drivetrain dynamics simulations need to be run under various wind load scenarios in the RBDO process. For this reason, a numerical procedure based on the multivariable tabular contact search algorithm is applied to the modeling of wind turbine drivetrains to reduce the overall computational time while retaining the precise contact geometry required for considering the gear tooth profile optimization. An integrated computational framework for the wind turbine drivetrain RBDO is then developed by incorporating the wind load uncertainty, the rotor blade aerodynamics model, the drivetrain dynamics model, and the probabilistic contact fatigue failure model. It is demonstrated that the RBDO optimum for a 750 kW wind turbine drivetrain obtained using the procedure developed in this study can achieve the target 97.725% reliability (2 sigma quality level) with only a 1.4% increase in the total weight from the baseline design, which had a reliability of 8.3%. Furthermore, it is shown that the tooth profile optimization, tip relief introduced as a design variable, prevents a large increase of the face width that would result in a large increase in the weight (cost) of the drivetrain in order to satisfy the target reliability against the tooth contact fatigue failure.     

风电机组齿轮箱的多变量时间序列故障预警

针对风电机组故障预警中，原始动态时间规整（DTW）算法无法有效度量风电机组多变量时间序列数据之间距离的问题，提出一种基于犹豫模糊集的动态时间规整（HFS-DTW）算法。该算法是原始DTW算法的一种扩展算法，可对单变量和多变量时间序列数据进行距离度量，且精度与速度较原始DTW算法更优。以子时间序列相似度距离为目标函数，使用帝国竞争算法（ICA）优化了HFS-DTW算法中的子序列长度和步距参数。算例研究表明与仅DTW算法和非参数最优的HFS-DTW算法相对比，参数最优的HFS-DTW可挖掘更多的多维特征点信息，输出的多维特征点相似序列具有更丰富细节；且基于所提算法可提前10天预警风电机组齿轮箱故障。     

分形插值在风速时间序列中的应用

针对风电场风速数据中大量连续缺失数据的插值问题,提出了一种基于自适应变异粒子群优化（PSO）的分形插值算法。首先,在粒子群优化算法中引入变异因子,增强粒子的多样性,提高算法搜索精度;其次,通过自适应变异粒子群优化算法来得到分形插值算法中垂直比例因子参数的最佳取值;最后,对两组不同趋势和变化特征的数据集进行分形插值计算分析,并把所提算法与Lagrange插值和三次样条插值方法进行对比。结果表明：分形插值不仅可以保持风速曲线的整体波动特性和局部特征,而且比传统插值方法的精度更高;在基于Dataset A的实验中,分形插值的均方根误差（RMSE）分别比Lagrange插值和三次样条插值减小了66.52%和58.57%;在基于Dataset B的实验中,分形插值的RMSE分别比Lagrange插值和三次样条插值减小了76.72%和67.33%。证明分形插值更适合连续缺失且波动强烈的风速时间序列的插值。     

分形插值在风速时间序列中的应用

针对风电场风速数据中大量连续缺失数据的插值问题，提出了一种基于自适应变异粒子群优化（PSO）的分形插值算法。首先，在粒子群优化算法中引入变异因子，增强粒子的多样性，提高算法搜索精度；其次，通过自适应变异粒子群优化算法来得到分形插值算法中垂直比例因子参数的最佳取值；最后，对两组不同趋势和变化特征的数据集进行分形插值计算分析，并把所提算法与Lagrange插值和三次样条插值方法进行对比。结果表明：分形插值不仅可以保持风速曲线的整体波动特性和局部特征，而且比传统插值方法的精度更高；在基于Dataset A的实验中，分形插值的均方根误差（RMSE）分别比Lagrange插值和三次样条插值减小了66.52%和58.57%；在基于Dataset B的实验中，分形插值的RMSE分别比Lagrange插值和三次样条插值减小了76.72%和67.33%。证明分形插值更适合连续缺失且波动强烈的风速时间序列的插值。     

Global data-driven modeling of wind turbines in the presence of turbulence

This paper presents a practical approach to identify a global model of a wind turbine from operational data, while it operates in a turbulent wind field with a varying mean wind speed and under closed-loop control. The approach is based on the realization that the nonlinearities are dominated by the aerodynamics of the rotor, which change with the operating condition. The dynamics of a wind turbine can be decomposed into a nonlinear static part, governed by the torque and thrust characteristics of the rotor, and a linear time-invariant dynamic part. The multi-input–multi-output linear dynamics are estimated using a recent closed-loop subspace identification method. The practical applicability of the algorithm is demonstrated by applying it to data obtained from the NREL CART 3 research turbine.     

An analytical algorithm with a wave age factor for altimeter wind speed retrieval

Based on the specular reflection theory of electromagnetic waves at rough sea surface and the wind wave spectrum model with a wave age factor, the sea surface wind speeds are retrieved from the normalized radar backscatter cross‐section (NRCS) measured by TOPEX/Poseidon (T/P) Ku‐band altimeter using the mean square slope (MSS) calculated from the spectrum models of the wind waves and the gravity‐capillary waves. A relationship between wave age and non‐dimensional wave height is applied to compute the wave age factor using the significant wave height (SWH) and wind speeds obtained from buoy or altimeter simultaneously. The study indicates that the wave age factor has a significant impact on the retrieval of altimeter wind speed. Compared with the operational algorithm for retrieving altimeter wind speed, the wind speed retrieved from the new analytical algorithm based on the wind wave spectrum model with the wave age factor, proposed in this study, can match the buoy measurements better. The effects of the wave age factor on altimeter wind speed retrieval are also shown quantitatively through a series of experiments and measurements. The comparison with the operational algorithm indicates that both the bias and root mean square error (RMSE) between wind speeds retrieved by the proposed analytical algorithm and those observed by the buoy decrease significantly. In the Gulf of Mexico, with the new analytical algorithm, more accurate altimeter wind speeds are retrieved.     

Computer simulation of buffeting actions of suspension bridges under turbulent wind

Suspension bridges are long-span flexible structures susceptible to various types of wind induced vibrations such as buffeting actions. In this paper, a three dimensional finite-element model formulated to deal with suspension bridges under turbulent wind is presented. In this model, all sources of geometric nonlinearity such as cable sag, force-bending moment interaction in the bridge deck and towers, and changes of bridge geometry due to large displacements, are fully considered. The wind loads, composed of steady-state wind loads, buffeting loads and self-excited loads, are converted into time domain by using the computer simulation technique. The Newmark-β step by step numerical integration algorithm is used to calculate the buffeting responses of bridges. Compared with the results obtained by classical buffeting theory, the validity of the simulation is proved.     

High-resolution satellite measurements of coastal wind field and sea state

Methods to derive wind speed and sea state by simple empirical models from synthetic aperture radar (SAR) data are presented and applied for use in high-resolution numerical modelling for coastal applications. The new radar satellite, TerraSAR-X (TS-X), images the surface of the sea with a high resolution up to 1 m. Therefore, not only wind information and integrated sea state parameters but also individual ocean waves with wavelengths down to 30 m are detectable. Two-dimensional information on the ocean surface retrieved using TS-X data is validated for different oceanographic applications: derivation of finely resolved wind fields (XMOD algorithm) and integrated sea state parameters (XWAVE algorithm). Both algorithms are capable of taking into account fine-scale effects in coastal areas. Wind and sea state information retrieved from SAR data are applied as the input for a wave numerical spectral model (wind forcing and boundary condition) running at a fine spatial horizontal resolution of 100 m. Results are compared to collocated buoy measurements. Studies are carried out for varying wind speeds and comparisons against wave height, simulated using original TS-X-derived wind data, showing the sensitivity of waves to local wind variation and thus the importance of local wind effects on wave behaviour in coastal areas. Examples for the German Bight (North Sea) are shown. The TS-X satellite scenes render well-developed ocean wave patterns of developed swell at the sea surface. Refraction of individual long swell waves at a water depth shallower than about 70 m, caused by the influence of underwater topography in coastal areas, is imaged on the radar scenes. A technique is developed for tracking wave rays depending on changes in swell wavelength and direction. We estimate the wave energy flux along wave tracks from deep water to the coastline based on SAR information: wave height and wavelength are derived from TS-X data.     

动态聚类适应值共享遗传算法在散射计风矢量反演中的应用

风矢量反演是散射计数据处理的核心内容,传统风矢量反演算法的设计过多依赖于目标函数的具体分布形态。以SeaWinds散射计为例,根据风矢量反演的多解问题和模糊解特性,设计了一种基于动态小生境遗传算法的风矢量反演算法。利用部分L2A和相应L2B数据对该算法进行了验证。结果表明该算法在无需任何目标函数先验知识的条件下能够取得较好的反演结果。     

基于模糊C均值的低空风切变预警算法

针对新一代多普勒天气雷达CINRAD在径向或切向检测风切变时容易丢失小切变的问题,提出了一种基于模糊C均值（FCM）的低空风切变预警算法用于阵风锋和龙卷风引起的风切变识别中。该算法的核心思想是运用8邻域系统,根据风速梯度识别不同程度切变,从而实现高切变及低切变预警。首先,采用全变分（TV）模型对雷达速度基数据进行去噪,同时保持速度基数据的细节特征;其次,采用每个速度基数据及其8邻域系统分别对应的速度值依次与4个方向模板卷积,获取4个方位速度梯度值;然后,采用FCM算法将梯度值分为高低两类,实现不同强度的风切变预警。采用武汉暴雨研究所提供的实测基数据进行测试和验证,能较为准确地识别出小切变。实验结果表明,该算法检测出来的风切变在定位精度和边缘识别两个方面均优于基于径向或切向的风切变识别算法,这对判断风切变的位置和强度以及分析不同天气引起的风切变具有重要指导意义。     

On the seasonal inconsistency of altimeter wind speed algorithms

Altimeter-derived wind speed has become a major source of data for climatological studies in recent years, thanks to the continued success of several operational missions. A variety of algorithms have been proposed to convert altimeter-measured radar cross-section to sea surface wind speed. The performance of these algorithms determines, to a large extent, the quality of the estimated wind speeds and, consequently, their suitability for various applications. In this study, 6 years of coincident TOPEX altimeter and JMA (Japan Meteorological Agency) buoy wind speeds from January 1993 to December 1998 were used to evaluate the seasonal consistency of three algorithms. Our results demonstrate that seasonal inconsistency is a common problem associated with altimeter wind speed algorithms. A biased pattern of seasonal wind speed variation is inevitable as long as available altimeter algorithms are used in their present forms. The type and extent of the distortion depend on specific algorithm and actual wind speed distribution. It is shown that such a systematic error may have a number of potential consequences, which could undermine some of the geophysical interpretations concerning the seasonality of global and regional wind climate.     

差分进化法在风电并网系统中的应用

为解决多机风电并网系统的稳定性问题,提出在风电并网系统的同步发电机(SG)中安装电力系统稳定器(PSS),利用差分进化算法解决SG中自动电压调节器(AVR)和PSS参数的最优调节问题。在有、无PSS以及是否使用差分进化法的各种情况下对风电并网系统稳定性进行了研究分析,研究表明通过差分进化法的协同调节使含AVR和PSS的风电并网系统有良好的阻尼作用,能减少发电机转子角差振荡,提高电压稳定性,通过仿真结果对比可知差分进化法可使并网系统稳定性明显提高。     

基于大数据分析的风机叶片结冰故障诊断

为了实现风力发电机叶片结冰故障诊断,及时进行风机叶片除冰,消除隐患。提出了基于大数据分析的人工智能算法识别风机叶片结冰的方法。首先,用结冰机理研究和数据探索的方法对风机运行数据进行分析,初步提取了24个特征量;然后,采用遗传算法对24个特征量、滑动窗口宽度和支持向量机参数进行联合优化,并据此建立叶片结冰故障诊断模型。诊断结果表明,用该模型诊断叶片结冰故障的准确率为86.2%,比采用SCADA采集所有数据或初步提取的24个特征量作为模型输入的准确率有大幅度的提高;并且,将该模型用于另一个#2风机时,故障诊断准确率也达到了78.5%,证明了该方法的有效性,并具有较好的泛化能力,为识别风机叶片结冰故障提供了新思路。     

Evolutionary product unit neural networks for short-term wind speed forecasting in wind farms

Combinations of physical and statistical wind speed forecasting models are frequently used in wind speed prediction problems arising in wind farms management. Artificial neural networks can be used in these models as a final step to obtain accurate wind speed predictions. The aim of this work is to determine the potential of evolutionary product unit neural networks (EPUNNs) for improving the accuracy and interpretation of these systems. Traditional neural network and EPUNN approaches have been used to develop different wind speed prediction models. The results obtained using different EPUNN models show that the functional model and the hybrid algorithms proposed provide very accurate prediction compared with standard neural networks used to solve this regression problem. One of the main advantages of the application of these EPUNNs has been the possibility of obtaining some interpretation of the non-linear relation predicted by the model, as will be shown in real data of a wind farm in Spain.     

Fusion of sea surface wind vector data acquired by multi-source active and passive sensors in China sea

This work is the first to analyse the sea surface wind vector (SSWV) data acquisition capabilities of eight satellites carrying microwave scatterometer (scanning scatterometer carried by Haiyang satellite 2A, advanced scatterometer carried by Metop satellite A, advanced scatterometer carried by Metop satellite B and scanning scatterometer carried by Oceansat satellite 2) or radiometers (Special Sensor Microwave Imager carried by Meteorological Satellite Program satellites F15 and F17, advanced microwave scanning radiometer 2 carried by GCOM-W1 satellite, and windsat polarimetric radiometer carried by Coriolis satellite) and investigate a SSWV fusion algorithm for active and passive remote-sensing data. We found that combining observations of the eight satellites can provide an SSWV data product with spatial resolution of 25 km × 25 km and temporal resolution of 3 h. Sea surface wind speed and direction data were obtained from multi-source active and passive sensors using a spatiotemporally weighted fusion algorithm. An adaptive sliding window was introduced for calculating effective observation data within spatial/temporal radii, which can effectively improve calculation efficiency for wind field fusion. Comparing the fused and buoy observation results, the root-mean-square errors of the wind direction and speed were 20.6° and 1.2 m s^–1, respectively, indicating that the fusion results can meet most application requirements for wind vector. Meanwhile, the space coverage, accuracy of merged wind speed and wind direction can be improved comparing to a single sensor.     

一种改进多亲遗传算法的并行模型研究

通过分析传统遗传算法和多亲遗传算法的不足,提出了一种多亲遗传算法的改进算法：基于共享存储器的多亲遗传算法,并对其进行了理论分析,讨论了GA的并行模型特点后,结合粗粒度并行模型和群体分组的并行方式,提出了一种MGASM的并行模型,该模型有利于改进MGASM的性能,提高其搜索效率。将MGASM-PPGA应用到了数据聚类问题中,进行了仿真实验,获得了理想的实验结果。     

风电系统最大风能追踪的智能模型预测控制

根据最大风能捕获原理,额定风速以下风能的最大追踪可以通过控制双馈感应发电机（DFIG）跟踪最优转速来实现。以变速恒频双馈风力发电系统为研究对象,研究了额定风速以下风能的最大追踪控制问题。首先针对双馈发电机强耦合、强非线性、机理模型复杂的特点,采用支持向量机（SVM）理论建立了智能预测模型;然后利用反馈校正的方法对预测输出进行修正,构成控制闭环;最后利用粒子群优化算法（PSO）调整参数少、演化群体小、计算速度快的优点容易地求出最优控制序列,较好地解决了滚动优化计算中的“瓶颈问题”。仿真结果验证了所采用的预测模型具有比较好的抗干扰能力和泛化能力,预测控制算法能够实现控制目标。     

Using differential evolution for fine tuning naïve Bayesian classifiers and its application for text classification

The Naive Bayes (NB) learning algorithm is simple and effective in many domains including text classification. However, its performance depends on the accuracy of the estimated conditional probability terms. Sometimes these terms are hard to be accurately estimated especially when the training data is scarce. This work transforms the probability estimation problem into an optimization problem, and exploits three metaheuristic approaches to solve it. These approaches are Genetic Algorithms (GA), Simulated Annealing (SA), and Differential Evolution (DE). We also propose a novel DE algorithm that uses multi-parent mutation and crossover operations (MPDE) and three different methods to select the final solution. We create an initial population by manipulating the solution generated by a method used for fine tuning the NB. We evaluate the proposed methods by using their resulted solutions to build NB classifiers and compare their results with the results of obtained from classical NB and Fine-Tuning Naïve Bayesian (FTNB) algorithm, using 53 UCI benchmark data sets. We name these obtained classifiers NBGA, NBSA, NBDE, and NB-MPDE respectively. We also evaluate the performance NB-MPDE for text-classification using 18 text-classification data sets, and compare its results with the results of obtained from FTNB, BNB, and MNB. The experimental results show that using DE in general and the proposed MPDE algorithm in particular are more convenient for fine-tuning NB than all other methods, including the other two metaheuristic methods (GA, and SA). They also indicate that NB-MPDE achieves superiority over classical NB, FTNB, NBDE, NBGA, NBSA, MNB, and BNB.