论文部分内容阅读
电力系统负荷聚类是大区电网负荷建模的基础工作之一,文中提出了一种基于粒子群优化的并行神经网络的电力系统负荷聚类算法。为了增加网络的并行处理能力,分别用一定数量的子样本集轮流对一定数量的神经网络进行并行训练,训练的结果再经过粒子群的优化,最终得到一个最优的聚类神经网络;同时为了克服神经网络聚类算法对输入样本的敏感性问题,算法采用非线性的连接权函数并将其中心作为粒子;给出了算法实现过程。采用东北电网负荷模型统计样本数据的聚类结果表明,文中提出的算法具有较强的适应性和较好的效果。
Power system load clustering is one of the basic tasks of load modeling in power grids. In this paper, a power system load clustering algorithm based on PSO and parallel neural networks is proposed. In order to increase the parallel processing ability of the network, a certain number of sub-sample sets are used in parallel to train a certain number of neural networks in turn. The training results are optimized by particle swarm optimization to obtain an optimal clustering neural network. It overcomes the sensitivity of the neural network clustering algorithm to the input samples. The algorithm uses the non-linear connection weight function and takes the center as the particle. The algorithm realization process is given. The clustering results of statistical sample data of load model of Northeast Power Grid show that the proposed algorithm has strong adaptability and good effect.