基于多变量时空融合网络的风机数据缺失值插补研究

风电场数据的完整性会因恶劣天气、输入信号丢失、传感器故障等原因遭到破坏,而大面积的数据缺失将给风机设备的运行和维护带来严峻考验.因此,提出一个多变量时空融合网络(Multivariate spatiotemporal integration network,MSIN)来解决缺失数据问题.首先,提出包含缺失值定位−指引机制的MSIN结构,揭示缺失部分数据的潜在信息,确保插补数据符合真实分布.其次,在网络中设计多视角时空卷积模块,捕捉同一风机多个变量与多个风机同一变量之间的局部空间和全局时间相关性,用于提高插补数据的真实性.接着,提出网络实时自更新机制,根据风电场实时变化情况实现在线调整,能够提升网络泛化能力,由此弥补重新训练模型的时间和空间成本高的缺陷.最后,通过真实的风机数据验证所提网络的有效性和优越性.相关分析结果表明,相较于MissForest等传统数据插补方法的插补性能,平均绝对误差(Mean absolute error,MAE)、平均绝对百分比误差(Mean absolute percentage error,MAPE)和均方根误差(Root mean square error,RMSE)分别下降18.54%、41.00%和3.15%以上.

面向新能源发电的风机数据智能分析订正技术研究

针对现有风速预测手段有限且准确率较低的问题,文中提出了一种面向新能源发电的风机数据智能分析订正模型。该模型将风速、风向和气压等元素的历史数据作为输入,并利用改进的随机森林算法(RF)完成对重要元素的筛选,从而降低样本空间的维度。同时通过长短期记忆网络进行重要元素与风速之间内在关联关系的挖掘,实现精准的风速数据预测及订正分析。算法分析结果表明,所设计的RF-LSTM算法相较于ECMWF方案,风向误差下降了约40°,风速误差则降低了0.5 m/s以上,预测准确率显著提升。

Wind Energy Potential Assessment in Republic of Macedonia

The paper presents the measurement campaign of wind energy potential undertaken in Republic of Macedonia on four sites from the middle of 2006. The wind data analysis has been performed for one site, following with the assessment of energy production of simulated wind park with six wind turbine generators.

Experimental Investigation of Reynolds Number Effect on Wind Turbine Profiles in the Cryogenic Wind Tunnel Cologne DNW-KKK

Wind turbine size has increased continuously and correspondingly also its Reynolds numbers. The Reynolds number effect can therefore no longer be ignored in design and optimization of wind turbines. Reliable profile test data should be available. A suitable facility for testing wind turbine profiles at high Reynolds numbers is the Cryogenic Wind Tunnel Cologne DNW-KKK. By means of injecting liquid nitrogen the tunnel can be cooled down to 100 K and the Reynolds number therefore can be raised accordingly. The maximum Reynolds number for 2D profile tests can reach 27x10^6. In this paper the test uncertainty and the flow quality of DNW-KKK were analyzed. Then some test results on the Reynolds number effect of the wind turbine profiles will be presented. The Reynolds number effect is different from model to model. Especially for thick profiles and flow control devices the Reynolds number effect is not always like the description in literature.

Numerical Simulation on Ice Shedding Phenomena in Turbomachinery

In the jet engine, icing phenomena occur primarily on the fan blades, the FEGVs （fan exit guide vanes）, the splitter, and the low-pressure compressor. Accreted ice disturbs the inlet flow and causes large energy losses. In addition, ice accreted on a fan rotor can be shed from the blade surface due to centrifugal force and can damage compressor components. This phenomenon, which is typical in turbomachinery, is referred to as ice shedding. Although existing icing models can simulate ice growth, these models do not have the capability to reproduce ice shedding. In the present study, we develop an icing model that takes into account both ice growth and ice shedding. Furthermore, we have validated the proposed ice shedding model through the comparison of numerical results and experimental data, which include the flow rate loss due to ice growth and the flow rate recovery due to ice shedding. The simulation results for the time at which ice shedding occurred and what were obtained using the proposed ice shedding model were in good agreement with the experimental results.

Hazard Model Reliability Analysis Based on a Wind Generator Condition Monitoring System

This paper presents an application of the hazard model reliability analysis on wind generators, based on a condition monitoring system. The hazard model techniques are most widely used in the statistical analysis of the electric machine＇s lifetime data. The model can be utilized to perform appropriate maintenance decision-making based on the evaluation of the mean time to failures that occur on the wind generators due to high temperatures. The knowledge of the condition monitoring system is used to estimate the hazard failure, and survival rates, which allows the preventive maintenance approach to be performed accurately. A case study is presented to demonstrate the adequacy of the proposed method based on the condition monitoring data for two wind turbines. Such data are representative in the generator temperatures with respect to the expended operating hours of the selected wind turbines. In this context, the influence of the generator temperatures on the lifetime of the generators can be determined. The results of the study can be used to develop the predetermined maintenance program, which significantly reduces the maintenance and operation costs.

Building Effects on a Horizontal-Axis Micro Wind Turbine： Experimental and Fluid-Dynamic Analysis

In this work, the efficiency ofa 1 kWp horizontal-axis wind turbine which is installed on the roof of the engineering building at the University of Salento has been evaluated, by means of CFD （computational fluid dynamic） and experimental data. Particularly, the influence of the building on the micro wind turbine performance has been studied and the numerical results （wind velocity fields and turbulence intensity above the building） have been compared with the experimental data collected over a period of three years. The results have shown that horizontal-axis wind turbines suffer from wake effect due to buildings, therefore, best sites in urban area have to be identified by a careful fluid dynamic analysis aimed at evaluating all causes that can reduce significantly the performance of the generator： in fact, building should allow to exploit increased wind intensity, but often this advantage is voided by turbulence phenomena, as in the case under investigation where the measured aerogenerator efficiency is lower than the nominal performance curve. Then, the best site can be found by crossing the contours of wind velocity with the turbulence intensity fields： in this way it is possible to localize an area （best location） where the aerogenerator can give maximum performance.

Air temperature feedback and its contribution to global warming

Air temperature feedback results from the thermal-radiative coupling between the atmosphere and the surface and plays an important role in surface energy balance. This paper reveals the contribution of air temperature feedback to the global warming from 1980 to 2000. The air temperature feedback kernel, evaluated using the ERA-Interim reanalysis data, is used to discuss the physical mechanism for air temperature feedback, the dependency of the strength of air temperature feedback on the climatological spatial distributions of air temperature, water vapor and cloud content, and the contributions of air temperature feedback to rapid global warming. The coupling between temperature feedback and each of the external forcings and individual feedback processes will amplify the anomaly of direct energy flux convergence at the surface induced by the external forcings and individual processes. The air temperature feedback amplifies the initial surface warming due to the increase in CO2 concentration, ice and snow melting, increase in water vapor, and change in ocean heat storage. It also amplifies the surface warming due to the longwave radiaitve forcing associated with the increase in cloud cover, which acts to suppress the cooling of the shortwave effect of cloud forcing. Overall, temperature feedback plays an important role in the global warming from 1980 to2000, as the net positive contribution to the perturbation of global mean energy flux at the surface from the air temperature feedback is larger than the net negative contribution from external forcing and all non-temperature feedbacks.

Design of Off-statistics Axial-flow Fans by Means of Vortex Law Optimization

Off-statistics input data sets are common in axial-flow fans design and may easily result in some violation of the requirements of a good aerodynamic blade design.In order to circumvent this problem,in the present paper,a solution to the radial equilibrium equation is found which minimizes the outlet kinetic energy and fulfills the aerodynamic constraints,thus ensuring that the resulting blade has acceptable aerodynamic performance.The presented method is based on the optimization of a three-parameters vortex law and of the meridional channel size.The aerodynamic quantities to be employed as constraints are individuated and their suitable ranges of variation are proposed.The method is validated by means of a design with critical input data values and CFD analysis.Then,by means of systematic computations with different input data sets,some correlations and charts are obtained which are analogous to classic correlations based on statistical investigations on existing machines.Such new correlations help size a fan of given characteristics as well as study the feasibility of a given design.