基于优化核极限学习机的中期电力负荷预测

中文摘要：

针对电力负荷预测，提出了一种优化的核极限学习机（O-KELM）的方法。核极限学习机（KELM）方法仅以核函数表示未知的隐含层非线性特征映射，无需选择隐含层的节点数目，通过正则化最小二乘算法计算网络的输出权值。将优化算法应用于KELM方法中，给出基于遗传算法、微分演化、模拟退火的3种优化KELM方法，优化选择核函数的参数以及正则化系数，以进一步提高KELM方法的学习性能。为验证方法的有效性，将O-KELM方法应用于某地区的中期峰值电力负荷预测研究中，在同等条件下与优化极限学习机（O-ELM）方法、SVM等方法进行比较。实验结果表明，O-KELM方法具有很好的预测性能，其中GA-KELM方法的建模精度最高。

Abstract:

An optimized kernel extreme learning machine(O-KELM) method is proposed for electricity load forecasting.Kernel extreme learning machine(KELM) method only uses the kernel function to represent the unknown nonlinear feature map of hidden layer,and does not need to choose the number of nodes in hidden layer,and calculates output weights of the network through the regularized least squares algorithm.The optimization algorithm was applied to the KELM method,and three optimized methods based on GA(genetic algorithm),DE(differential evolution) and SA(simulated annealing)were given to select the kernel function parameters and the regularization coefficients to further improve the learning performance of the KELM method.To verify the validity of the employed method,the O-KELM method was applied to the midterm electricity peak load forecasting in a region,and under the same conditions,it was compared with the methods of optimized extreme learning machine (O-ELM) and SVM.The experimental results show that the O-KELM method has good forecasting performance,and the GA-KELM method has the highest modeling accuracy.