摘要
提出一种新的基于结合PSO(Particle Swarm Optimization)和BP(Back Propagation)神经网络的优化算法,按季节、日类型划分12个子网络组成的预测模型,并以影响发电量的关键因素太阳辐射强度、气温、历史发电量作为输入变量,预测光伏电站日发电量。预测结果显示:该预测模型能保证在日类型等条件发生转变时模型的持续有效性,预测误差均小于20%,预测精度能满足电网公司要求。
A new optimization algorithm which based on combining PSO （particle swarm optimization）and BP （Back Propagation） neural network was proposed. The forecasting model was divided into twelve sub-networks according to the type of season and day. Irradiance, temperature and historical electricity, which are the key factors to affect generated electrical energy, were treated as input variables of predicted model. The forecasting results showed that the predicted model can ensure continuing effectiveness of forecasting when weather patterns transform, the maximum error of the predicted model for the photovoltaic power stations in all the prediction time was no more than 20%, and the forecasting accuracy can satisfy the requirements of power grid corporation.
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2016年第1期63-68,共6页
Acta Energiae Solaris Sinica
基金
湖北省经济信息委员会资助项目"太阳能光伏并网及智能化调度一体化研究"
关键词
光伏发电
粒子群优化算法(PSO)
神经网络
预测精度
photovoltaic power generation
particle swarm optimization（PSO）
neural network
forecasting accuracy
