一种基于粒子群优化并行神经网络的电力系统负荷特性聚类方法

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。

Characteristics Clustering Approach of Power System Load Based on Parallel Neural Network with Particle Swarm Optimization

Characteristics clustering of power system load is one of the key works in load modeling of large-scale power grid. This paper presents a characteristics clustering approach of power system load based on Parallel Neural Network (PNN) with Particle Swarm Optimization (PSO). In order to improve the parallel process ability of NN, the parallel training strategy is adopted in this approach, that is, the given number sub-sample datum are used to train the specific number NN by turns, then, the results of parallel training is optimized by PSO.Finally, an optimization clustering NN is gained. At the same time, in order to overcome the sensitivity of sample datum, a nonlinear connection function is adopted in this neural network clustering algorithm, the centre of nonlinear connection function is regarded as a particle. The North-East China load model Statistic datum has been used to illustrate that the proposed algorithm on the characteristics clustering of power system load is effective.