基于自组织映射的改进 BP 神经网络短期光伏出力预测研究

Study on Prediction of Short － term Photovoltaic Output Power with Improved BP Neural Network Based on Self － organizing Mapping

摘要:

针对光伏发电出力随机波动给电网调度造成困难这一问题，提出了一种基于 SOM － PSO － BP 的模型对光伏 有功功率进行短期预测，用于提高电网对可再生能源的调度能力。首先采用自组织映射对原始数据组进行聚类降 维; 接着使用粒子群算法对 BP 神经网络的权重和偏置矩阵进行寻优; 然后利用训练集构造 SOM － PSO － BP 预测模 型; 最后在对比仿真中验证了所提方法的有效性。

Abstract:

The random fluctuation of photovoltaic power generation may cause difficulties for power grid dispatching，so a short － term prediction model based on SOM － PSO － BP is proposed，which will improve the dispatching ability of renewable energy by power grid． Firstly，the self － organizing mapping is used to reduce and cluster the dimension of the original data． Secondly，the weight and bias matrix of BP neural network are optimized by using particle swarm algorithm，and then the SOM － PSO － BP prediction model is constructed by using the training sets． Finally，the effectiveness of the proposed method is verified in the simulation．