摘要
大量随机冲击负荷(炼钢厂、轧钢厂)接入系统给负荷预测造成困难,使得短期负荷预测精度下降。首先就冲击负荷地区的负荷特征进行分析,找出影响负荷预测精度的原因,提出利用小波变换对负荷序列进行多尺度分解,得到在不同频段下负荷子序列,重点针对各负荷分量不同特点,建立含不同输入量的贝叶斯神经网络预测模型,再将预测结果进行小波重构,从而得出最后预测结果。再与另两种模型进行对比,结果表明该方法能提高冲击负荷地区的短期负荷预测精度。
A large number of random impact loads （ steel plant and steel-rolling plant） access systems create diffi- culties for load forecasting, decline in short-term load forecasting precision. This paper analyzes load characteristic of impact loads first; finds the cause affecting the precision of load forecasting, using muhiscale decomposition by wavelet transform of load sequence; obtains load subsequences under different band, with emphasis on different features for each load sequence, built different input Bayesian neural network forecasting models and forecast re- sults for the wavelet reconstruction, get the final prediction results. Then compared with other two models, results show that this method can improve the impact load area for short term load forecasting accuracy.
出处
《电力科学与工程》
2012年第11期7-12,共6页
Electric Power Science and Engineering
基金
广东省自然科学基金项目(10151009001000045)
南方电网公司2011年科技项目(YX-GD-296)
关键词
小波分析
贝叶斯神经网络
冲击负荷
短期负荷预测
wavelet analysis
Bayesian neural networks
impact load
short term load forecasting
