基于最小一乘法的组合赋权法在中长期负荷预测中的应用

中长期负荷预测是电网规划的重要依据和前提,出现了大量的算法模型,但每种模型都有自己的适用条件,采用组合预测能有效地组织各种模型做到扬长避短。组合预测法的关键是确定组合模型的权重,最小二乘法是应用较为广泛的确定模型权重的方法,但研究表明,该方法在确定负荷有突变情况的权重时存在较大的误差,因此,提出基于最小一乘法的权重确定方法,与最小二乘法相比,该方法以误差绝对值之和最小为优化目标,而非平方和最小,避免了误差的缩放,残差可以真实反映与真值的偏离程度,有利于权重的确定,最后通过实例验证了该方法的有效性和优越性。

Application of combination determining weights method based on LAD to the power mid-long term load forecasting

Middle and long term load forecasting is an important base of transmission and distribution network planning,there are many algorithm models for long-term load forecasting.However,each model applies on some conditions.Combined forecasting can effective overcome the short of every model.The key of combined forecast method is how to fix the association weight.Least square method is used widely to confirm weights.Researches show that least-square method has big errors under the conditions of load sudden changi...