基于Spiking神经网络的光伏发电系统功率预测  

Photovoltaic system power forecasting based on Spiking neutral network

作　　者：卢怡含 王林[2] 晋飞[2] 刘忠辉[2] 唐敏[2] 

机构地区：[1]国网昌邑市供电公司,山东昌邑261300 [2]国网潍坊供电公司,山东潍坊261000

出　　处：《黑龙江电力》2016年第3期263-269,共7页Heilongjiang Electric Power

摘　　要：针对光伏发电系统输出功率的随机性,提出了一种基于Spiking神经网络的光伏发电功率预测模型,它采用精确脉冲时间的编码方式,能接近真实的生物神经系统,并具有强大的计算能力。考虑天气类型、太阳辐照强度、环境温度等主要影响因素,采用近似欧式距离选取相似日的方法,应用实际光伏发电系统的历史发电数据和气象数据对Spiking神经网络、BP神经网络和支持向量机三种预测模型进行测试和评估。预测结果与实测值的比较表明:Spiking神经网络模型相比于BP神经网络和支持向量机模型具有较高的预测精度和较强的适用性,可作为解决光伏发电系统功率预测可行方法之一。Aiming at the randomness of photovoltaic system output power,this paper proposed the forecasting model of PV power generation based on Spiking neural network. It is a model that uses coding method with computing capability and accurate pulse time,which is closer to the real biological nerve system. Considering the main influencing factors such as weather types,sunshine intensity,temperature etc.,the paper adopted the approximate Euclidean distance to select similar days,and the historical generation data and meteorological data of the practical PV system to test and evaluate three forecasting models,including Spiking neural network,BP neural network and support vector machine. The results of the comparison between the forecast and the actual measured values reveal that Spiking neural network model,compared with BP neural network model and support vector machine model,has relatively high forecast accuracy and robust applicability,and can provide an effective and feasible way to forecast the PV system power generation.

关 键 词：光伏系统 SPIKING神经网络 脉冲响应模型 Spikeprop算法 发电功率预测 

分 类 号：TM615.2[电气工程—电力系统及自动化]