在电力系统对功率预测提出更高要求的形势下,风电功率区间预测的方法已经逐渐成为新的热点.文章利用预测区间(PIs)的思想来估计风电场输出功率的不确定性.在优化区间预测目标函数的基础上,利用核极限学习机(KELM)学习速度快,泛化能力强的优点,提出一种基于KELM的风电功率区间预测模型.并使用改进后的蝙蝠算法(IBA)对模型的参数进行优化.为了克服BA易陷入局部最优的缺点,增加了其搜索时的多样性,并加入动态惯性权重,使其收敛速度更快.最后,用河北某风电场的数据进行实验仿真,与传统BP神经网络模型和BA-ELM模型对比分析,结果表明文章提出的预测方法具有速度快,精度高的优点.%This paper use the idea of predictive interval (PIs) to estimate the uncertainty of wind power. On the basis of optimizing the objective function of interval prediction, this paper proposes a KELM-based wind power interval prediction model taking the advantages of learning speed and extensive generalization ability. The parameters of the model were optimized by a modified bat algorithm (IBA). Due to the easy trap in local optima, the diversity of searching and dynamic inertia weight are added to make the bat algorithm(BA)converge faster. Finally, the simulation of a wind farm in Hebei Province is carried out. Compared with the traditional BP neural network model and the BA-ELM model, the results show that the proposed prediction method has the advantages of faster speed and higher precision.
展开▼