一种改进的基于分形插值和小波分析的电力负荷短期预测方法

An Improved Short-Term Load Forecasting Based on Fractal Interpretation and Wavelet Analysis

针对传统分形插值预测中关键参数垂直比例因子值计算困难的问题,提出一种基于分形插值曲线分维数定理和小波分析的改进垂直比例因子计算法,该算法通过对样本数据进行小波自相似性分析,利用自相似指数计算出分维数,间接求解出垂直比例因子值;然后就传统分形插值预测中插值点集合经验性选取的问题,提出了一种应用小波特征点检测法来选取插值点集合的方法。通过某电力公司原始负荷记录数据对某日24 h进行预测,并与经验分块法求得的预测值进行比较,算例结果表明,改进的分形插值预测法提高了预测精度,避免了分块法造成的迭代函数发散问题,且对样本需求量少、可操作性强。

In allusion to the difficulty in the calculation of vertical scale factors of key parameters in traditional load forecasting based on fractal interpolation, based on fractal dimension theorem of fractal interpolation curve and wavelet analysis an improved method to calculate vertical scale factors is proposed. In the proposed method, firstly by means of wavelet self-similarity analysis on sampled data the fractal dimension is calculated by self-similarity index, and the vertical scale factors are indirectly solved; then applying wavelet feature point detection, a method to choose interpolation point set is put forward. Using original load data of a certain power company, the 24 hours load forecasting of a certain day is performed, and the obtained results are compared with the results forecasted by empirical blocking method. Comparison result shows that using the proposed method the load forecasting accuracy is improved, the divergence of interpolation function caused by the blocking method can be avoided, and less sampled data is needed.