基于混合神经网络的风电场风资源评估  

Wind Energy Resource Assessment of Wind Farm Based on Hybrid Neural Network

作　　者：王娜[1] 周有庆[1] 邵霞[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《电工技术学报》2015年第14期370-376,共7页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金重点资助项目(51277055)

摘　　要：准确的风资源评估是风电场规划和设计的前提。为了提高风电场风资源评估的精度,提出了一种基于混合神经网络的风电场风资源评估方法,该方法可综合利用风电场附近区域信息进行评估。首先根据风电场和附近参考气象站的同期数据建立基于混合神经网络的相关模型,训练得到神经网络的权值参数,为了提高神经网络的学习能力和避免陷入局部最优,混合神经网络采用不同的训练方法,并且采用自适应粒子群算法进行优化;再将参考气象站的历史观测数据应用到该模型中,即可得到风电场的长期风速特性,在此基础上进行风资源评估参数的计算。仿真结果表明该方法具有较高的精度。Wind energy resource assessment is a key step of wind farm planning and design. A hybrid neural network（HNN） based wind energy resource assessment method is proposed to improve the assessment accuracy, and the method allows the use of regional information. Firstly, the HNN based correlate model is developed according to the concurrent wind speeds of the reference weather stations and the candidate wind farm. In order to obtain wider learning capability and avoid being trapped in a local minimum, the different training algorithm and the adaptive particle swarm optimization（APSO） are used in the HNN. Then, the whole long-term wind speed and direction data are applied to the model, thus the long-term wind characteristics of the candidate wind farm are obtained. The wind energy resource assessment parameters are subsequently computed on the basis of the knowledge of these wind speeds. The simulation results show that the proposed method has relatively high accuracy.

关 键 词：风电场 风资源评估 混合神经网络 自适应粒子群优化 

分 类 号：TM614[电气工程—电力系统及自动化]