摘要
提出以气象负荷和长期趋势负荷之和为聚类中心对历史负荷数据进行相似搜索的方法,该方法可提高预测样本同被预测日负荷的相似度,从而增加预测结果的可信度和精确度。以权重优化组合的方式采用多种负荷预测方法进行组合负荷预测,应用实例证明该方法可体现不同地区、不同类型、不同气象敏感度的负荷特性,因而具有广泛的自适应性,对于负荷总量较小、变动范围较大且受天气因素影响明显的地区具有较好的预测精度。
A similar search method for historical data, in which the sum of weather-connected load and long-term load are taken as clustering center, is proposed. This method can improve the similarity between the forecasting samples and the forecasted daily load, thus both confidence level and accuracy of forecasted results can be enhanced. By means of optimal combination of weights, the combinational load forecasting is conducted by various load forecasting methods. Application cases prove that the proposed method can incarnate the characteristics of different types of loads with different meteorological sensitivities in different regions, so this method possesses good adaptability and can provide better load forecasting accuracy for the regions where the total amounts of loads are not too big but vary in wide range and are easy impacted by whether factors.
出处
《电网技术》
EI
CSCD
北大核心
2007年第19期60-64,82,共6页
Power System Technology
基金
上海市教委科研项目(07ZZ145)
上海高校选拔培养优秀青年教师科研专项基金
上海市重点学科(P1301)
上海市科委重点项目(061612040)
关键词
负荷预测
气象因素
线性回归
时间序列
灰色模型
神经网络
组合预测
load forecasting
weather factors
linear regression
time series
grey model
neural network
combinational forecast