论文部分内容阅读
提高风电出力的预测精度,可以减轻风电并网带来的不利影响。利用径向基函数神经网络(RBF)建立风电出力预测模型,并通过正交二乘算法(OLS)对RBF神经网络进行初步训练,以确定网络结构及隐含层各节点中心。在OLS算法训练的网络基础上引入蛙跳算法(SFLA),进一步对隐含层基函数的宽度值进行优化以提高网络的泛化能力。实例预测表明,在相同的网络结构及隐含层中心下,基函数宽度值优化后的RBF神经网络模型预测精度得到了提升。
Improve the prediction accuracy of wind power output, can reduce the adverse effects of wind power grid. The radial basis function neural network (RBF) is used to establish the forecast model of wind power output, and the initial training of RBF neural network is carried out by the Orthogonal-Squares Algorithm (OLS) to determine the network structure and the nodes of hidden layer. Based on the network trained by OLS algorithm, frog leaping algorithm (SFLA) is introduced to further optimize the width of hidden layer function to improve the generalization ability of the network. The case study shows that under the same network structure and hidden layer center, the prediction accuracy of RBF neural network model with optimized basis width is improved.