自编码器融合极限学习机的广义负荷建模

General Load Modeling of Auto-Encoder Fused Extreme Learning Machine

随着各种分布式电源大规模并网,传统的负荷建模方法难以精确描述电力系统的负荷信息。为提高负荷区域的建模精度,广义负荷建模问题被提出。将机器学习理论引入广义负荷建模领域,提出一种基于自编码器融合极限学习机的广义负荷建模方法。首先,利用自编码器能降低输入数据维度的优势,提取特征值,通过其可最小化重构误差的特点,求得自编码器结构。然后,将此结构作为极限学习机的输入端结构,可得到已优化隐层节点数的极限学习机结构。最后,通过极限学习机的有监督学习方法,调整隐层至输出层的权值,保证网络收敛至最优值。搭建含有蓄电池和风力发电系统的广义负荷模型进行仿真测试。结果证明,该方法具有较高的建模精度,可以有效应用于含不同成分的电力系统广义负荷建模。

As different kinds of distributed power source are connected to the grid on a large scale,traditional load modeling methods are difficult to accurately describe the load information of the power system.In order to improve the modeling accuracy of the load region,general load modeling problem is raised.By introducing machine learning theory into general load modeling field,ageneral load modeling method based on auto-encoder fused extreme learning machine is proposed.Firstly,the advantages that the auto-encoder can reduce the dimension of the input data are used to extract the eigenvalues,and the structure of the auto-encoder is obtained by using its characteristics of minimizing the reconstruction error.Then,taking this structure as the input structure of the extreme learning machine,the structure of the extreme learning machine with the optimized hidden layer node number is obtained.Finally,the network is guaranteed to converge to the optimal value through adjusting the weight from hidden layer to output layer by the supervised learning method of the extreme learning machine.The general load model in power system containing battery and wind power source is built for simulation test.The results verify that the proposed method has high modeling accuracy and can be effectively applied to the general load modeling of the power system with different sources.