主分量启发式约简算法的短期负荷预测

Short-term Load Forecasting Based on Main Attribute-component Heuristic Algorithm

针对众多不确定要素影响电力负荷预测准确度的问题,对粗糙集理论进行研究的基础上,提出了基于属性主分量的启发式约简算法,对天气及负荷历史数据进行处理,并建立了与广义回归神经网络结合的短期负荷预测模型。通过属性约简算法提取对未知负荷变化影响大的关键要素,得到的要素属性作为该模型的学习样本。算例结果表明,新算法与根据经验选取输入参数的传统广义回归神经网络相比,预测准确度有了明显的提高,计算量减少,更适用于短期电力负荷预测。

Because of various factors that influence load forecasting accuracy, a reduction algorithm through main attribute-component algorithm based on rough set theory is introduced in this paper. To deal with the date of weath- er and history load data, and then establish a model combining with generalized regression neural network. The key factors influencing loads are performed by reduction algorithm, using them as the learning samples of generalized regression neural network. Forecasting results of calculation examples show that the forecasting accuracy is obvi- ously improved, and the amount of calculation is reduced comparing with traditional generalized regression neural network model which chooses input parameters in the light of experience. And this method is more suitable to short-term load forecasting.