摘要
用电负荷的预测是一个包含很多不确定因素的复杂问题,通常根据已有负荷数据建立模型进行预测,将近期负荷数据与远期负荷数据相结合,综合不同特征的负荷数据信息。针对单项预测模型在电网实时日负荷预测的局限性,建立了由ARMA模型和ANN模型为基础结合纵横交叉算法的混合预测模型。该模型采用时间序列模型处理连续纵向负荷数据,采用神经网络技术处理断续横向负荷数据,对计算得到的纵向、横向预测值加权计算,获得交叉预测值。提出的对纵向连续负荷数据和横向断续数据分别建模的纵横交叉预测法,以获取尽可能多的重要预测信息,提高预测精度。在某电网上进行了实例测试,具有较好的效果。
Power load forecasting is quite complex due to various uncertainties. Based on the forecasting model established by historical load data, the recent load data should be combined with the forward load data to synthesize the load data of different characteristics. Aiming at the limitations of single forecasting model in real-time daily load forecasting, a CSO-based hybrid forecasting model combining auto-regressive and moving average model（ARMA） and neural network algorithm（ANN） is established. The model proposes a time series model to deal with continuous longitudinal load data, and neural network technology to deal with intermittent lateral load data; the weighted calculation on the longitudinal and horizontal predictive is conducted to obtain cross prediction value. A vertical and horizontal cross prediction method for longitudinal continuous load data and transverse intermittent data is proposed to obtain important predictive information as much as possible to improve the prediction accuracy. The method is used in a power grid and achieves favorable effect.
作者
范金骥
FAN Jinji(Datang East China Electric Power Test ＆ Research Institute,Hefei 230000,China)
出处
《浙江电力》
2018年第8期35-41,共7页
Zhejiang Electric Power
