中文摘要：

提出了一种基于层次结构的中长期电力负荷变权组合预测方法，借鉴层次分析法的思想，构造一个层次结构来确定组合权重：采用熵值法确定模型评价指标的相对权重；采用方差一协方差优选组合预测方法和灰色关联分析分别确定各单一预测模型在各评价指标下的相对权重，最终确定组合预测模型中的组合权重。在组合预测的整个过程中，根据负荷发展的“近大远小”原则，引入等维信息的概念，实现了变权组合预测，使预测结果能够更合理地反映电力负荷的发展规律。最后通过一个实例验证了该方法的有效性。

英文摘要：

A new variable weight combination method for mid-long term power load forecasting based on hierarchical structure is presented. According to the idea of analytic hierarchy process, a hierarchical structure for determining combined weights was constructed： entropy method was introduced to ascertain the weights of model-evaluation indexes; variance-covariance optimization combination method and grey relational analysis were used to ascertain the weights of each single model under each evaluation index respectively, and determine the combined weights of combination forecasting model finally. In the whole process of combination forecasting, equal dimension information was introduced according to the ＂inertia principle＂ of load development, thus achieved variable weight combination forecasting, and forecasting results reasonably reflect the development law of power load. At last, the reliability was validated by an actual example.