摘要
为提高短期负荷预测精度,改善预测模型的工程实用性,提出了一种结合纵向和横向相似日样本的短期负荷二维组合预测方法。常用BP神经网络需优化的参数多,故采用结构简单的广义回归神经网络作为单向模型的基本预测算法。在此基础上,再通过设置组合加权系数,运用粒子群优化算法寻优系数值,得到最终的二维预测结果。对比其他短期尤其是超短期负荷预测方法,该模型不仅考虑了气象因素对负荷的影响,还充分体现了"近大远小"原则,并智能优化系数组合预测结果。电网实际负荷数据验证表明,该预测模型操作性高,速度快,且有较高的预测精度。
To improve the accuracy and engineering applicability of short-term load forecasting（STLF）model,a bidi-rectional（i.e.,longitudinal and transverse）combined STLF model is proposed. Considering that BP neural network hastoo many parameters to be determined,a generalized regression neural network which has a simple structure is adoptedas the basic forecasting method for a unidirectional model. On this basis,combined weight coefficients are set and opti-mized by particle swarm optimization algorithm to search for the optimization coefficients,and finally obtain the bidirec-tional forecasting results. Compared with other STLF methods,especially the ultra-STLF,the proposed model not onlytakes weather information into account,but also reflects the principle that the nearer data have a greater impact on theforecasting values. Moreover,it can optimize the coefficients intelligently. The actual grid load data indicate that theproposed forecasting model has easy operability,fast speed and higher forecasting accuracy.
出处
《电力系统及其自动化学报》
CSCD
北大核心
2018年第2期85-89,共5页
Proceedings of the CSU-EPSA
关键词
短期负荷预测
粒子群优化
神经网络
相似日
组合预测
short-term load forecasting
particle swarm optimization
neural network
similar day
combined forecasting
