Zhao P.,Xian Jiaotong University |
Xia J.,Xian Jiaotong University |
Dai Y.,Xian Jiaotong University |
He J.,Zhong Neng Power Technology Development Co.
Proceedings of the 2010 5th IEEE Conference on Industrial Electronics and Applications, ICIEA 2010 | Year: 2010
In this paper the wind speed forecasting in a wind farm, applying the algorithm of support vector regression (SVR) to the mean 10-minute time series is presented. By comparing its performance with an back propagation neural network model through simulation results, we could find following facts: firstly, both algorithms are applicable for prediction the wind speed time series in future; secondly, the prediction effect of support vector regression outperforms the back propagation neural network model as indicated by the prediction graph and by the mean square errors and mean absolute errors. Finally, we selected three different stages of the wind speed curve to analyze, the results show that the proposed algorithm fit the original wind speed curve well at the whole process, but the back propagation neural network is inapplicability for the rise stage when the ascent rate suddenly become flatness of the original wind speed curve. © 2010 IEEE.
Duan Y.,Auburn University |
Duan Y.,Zhong Neng Power Technology Development Co. |
Hosseinizadeh S.F.,Auburn University |
Hosseinizadeh S.F.,Sharif University of Technology |
And 2 more authors.
Journal of Heat Transfer | Year: 2010
The effects of insulated and isothermal thin baffles on pseudosteady-state natural convection within spherical containers were studied computationally. The computations are based on an iterative, finite-volume numerical procedure using primitive dependent variables. Natural convection effect is modeled via the Boussinesq approximation. Parametric studies were performed for a Prandtl number of 0.7. For Rayleigh numbers of 104, 105, 106, and 107, baffles with three lengths positioned at five different locations were investigated (120 cases). The fluid that is heated adjacent to the sphere rises replacing the colder fluid, which sinks downward through the stratified stable thermal layer. For high Ra number cases, the hot fluid at the bottom of the sphere is also observed to rise along the symmetry axis and encounter the sinking colder fluid, thus causing oscillations in the temperature and flow fields. Due to flow obstruction (blockage or confinement) effect of baffles and also because of the extra heating afforded by the isothermal baffle, multi-cell recirculating vortices are observed. This additional heat is directly linked to creation of another recirculating vortex next to the baffle. In effect, hot fluid is directed into the center of the sphere disrupting thermal stratified layers. For the majority of the baffles investigated, the Nusselt numbers were generally lower than the reference cases with no baffle. The extent of heat transfer modification depends on Ra, length, and location of the extended surface. With an insulated baffle, the lowest amount of absorbed heat corresponds to a baffle positioned horizontally. Placing a baffle near the top of the sphere for high Ra number cases can lead to heat transfer enhancement that is linked to disturbance of the thermal boundary layer. With isothermal baffles, heat transfer enhancement is achieved for a baffle placed near the bottom of the sphere due to interaction of the counterclockwise rotating vortex and the stratified layer. For some high Ra cases, strong fluctuations of the flow and thermal fields indicating departure from the pseudosteady-state were observed. © 2010 by ASME.
Wang M.,Zhong Neng Power Technology Development Co. |
Zeng L.,Hebei Electric Power Research Institute
Shuili Fadian Xuebao/Journal of Hydroelectric Engineering | Year: 2011
This paper discusses the existing models of wind speed frequency distribution such as log-normal distribution model, Weibull distribution model and AR-GARCH model, focusing on their features and limitations. An analysis of actual speed frequency distribution of a wind farm suggests that AR-GARCH model is the best in describing the actual wind. This model can fully reflect the randomness and wave character of wind speed that varies with time, and with a universality of application it effectively reduces the large fitting errors of the other models in low-speed and zero-speed zones, particularly for those distribution curves with multiple peaks. Application of the model to three typical engineering projects verified its prediction ability and accuracy. © Copyright.
Chen N.,Beihang University |
Qian Z.,Beihang University |
Meng X.,Beihang University |
Meng K.,Zhong Neng Power Technology Development Co.
Diangong Jishu Xuebao/Transactions of China Electrotechnical Society | Year: 2013
Accurate wind speed forecasting is necessary for evaluating the safety and economy of the large scale wind farm integration. In this paper, a new multi-step ahead wind speed forecasting model is presented based on spatial correlation and support vector machine (SVM) method. First, a wind direction oriented spatial correlation model is established, of which the optimized input vectors are determined by correlation coefficient. Then in order to eliminate the influence of variable wind direction, SVM method is applied to combine with the former spatial correlation model based on an accurate analysis of how forecast error depends on wind direction. The calculation results, which are obtained by measured data from a wind farm, indicate that the proposed spatial-SVM model has a better performance in forecasting accuracy comparing to the basic SVM model and other classical forecasting models.
Sun H.,Zhong Neng Power Technology Development Co. |
Li X.-W.,Xian Jiaotong University
Gaoya Dianqi/High Voltage Apparatus | Year: 2010
Chargeability is a crucial parameter for the design of ZnO arrester. For a 500 kV porcelain ZnO arrester, its potential distribution is calculated with finite element method, where reasonable boundary is set to simplify the solution of the open boundary field problem, and the strategy of coupled degrees of freedom is adopted to deal with the floating conductors. Then the potential distribution of the arrester is tested with optical fiber-current measurement method under the maximum continuous operating voltage. Compared with the calculation results, the validation of the proposed method is verified. Finally, the influences of the size of calculation boundary, the parameters of grading ring and the mounting height on potential distribution are analyzed, and the results demonstrate that the potential distribution coefficient of the upper and lower sections of MOA can be improved through increasing mounting height, and the lower grading ring has more significant influence on the potential distribution.