摘要
考虑到光伏发电系统输出功率特性,提出一种将经验分解法(EMD)与最小二乘支持向量机(LSSVM)相结合的方法对光伏发电系统功率进行预测。首先将历史数据按天气类型分类,利用欧氏距离挑选出待预测日的相似日数据;然后运用EMD将原始光伏发电系统功率序列分解为不同频率的相对平稳的IMF分量,将信号中存在的不同尺度波动或趋势逐级分解出来;最后对各IMF的每一时刻分别建立LSSVM预测模型,将各分量对应时刻的预测值等权值求和得到该时刻最终的光伏发电量。仿真预测结果表明,该方法与单一的LSSVM预测法及小波分解与LSSVM相结合预测法相比,预测精度得到大幅度的提高。
Considering the output power characteristics of PV power system, a prediction method based on combined method of empirical mode decomposition (EMD) and least square support vector machine (LSSVM) was proposed to predict output power of PV power systems. Firstly, the historical data were divided according to the weather conditions, and Euclidean distance was used to pick out the similar day data for the predicted date. Then the original power output data of PV systems were decomposed into a series intrinsic mode functions (IMF) by using EMD, so that different fluctuations and trends in the data could be decomposed progressively. Finally, LSSVM prediction model was established by every moment of each IMF, the PV power generation value should be the sum of the predicted value of each component corresponding to the time. The simulation predicted results show that the prediction accuracy has been greatly improved when compared with the single LSSVM prediction method or the combination prediction method of wavelet decomposition and LSSVM prediction method.
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2016年第6期1387-1395,共9页
Acta Energiae Solaris Sinica
基金
国家自然科学基金(51277057)