基于小波变换和支持向量机的短期光伏发电功率预测

Short-term Forecasting of Photovoltaic Power Generation Based on Wavelet Decomposition and Support Vector Machine

光伏发电功率预测是减小大规模光伏发电并网对电网造成不良影响的有效手段,对电网调度及光伏电站的优化运行具有重要意义。针对光伏发电功率序列的周期性和非平稳性,本文提出了基于小波变换和支持向量机(Support vector machine,SVM)的预测方法。文中对原始功率序列进行小波分解并单支重构,构成低频趋势信号和高频随机信号,利用具有小样本学习能力强和计算简单等特点的SVM对各小波数据序列分别预测,最终将各预测值合成得到预测功率值。某光伏发电站的实际数据仿真验证了该预测方法的可行性和有效性。

Photovoltaic power prediction is an effective way to reduce adverse effects caused by the large-scale photovoltaic power connected to grid, and it is of great significance for power grid scheduling and optimal operation of the photovoltaic power station. Considering the cyclical and non-stationary of photovoltaic power sequence, this paper provides a prediction method based on wavelet transform and support vector machine （SVM）. By wavelet decomposition and single refactoring, photovoltaic power sequence is converted to the low frequency trend signal and high frequency random signal. In consideration of strong small sample learning ability and small amount of calculation which SVM has, every wavelet signal are separately forecasted with support vector machine models. Finally, the predicted results of original photovoltaic power sequence are achieved by merging every single forecasted value. The actual data simulation validation of a photovoltaic power station shows the feasibility and effectiveness of this prediction method.