基于相似日的神经网络短期负荷预测方法

对于受不确定因素影响的短期电力负荷,提出了一种基于相似日的神经网络预测方法。设计了一个规范化的相关因素映射数据表,应用聚类分析方法描述由于相关因素的不同而导致的待预测日与历史日之间的差异程度,选用日特征量相同或相近的历史负荷数据作为神经网络的输入元素进行预测。用该方法选取相似日可以较多的考虑各种因素,因此,具有较高的预测精度。

Short-term load forecasting using Ann based on similar historical day data

For the short -term electric power load with uncertainty influence factors, we put forward the load forecasting method using ANN based on similar historical day. We design a standard mapping datasheet based on some correlated factors and apply clustering analysis method to describe the difference between character of the forecasting and historical day. This difference is caused by the difference of related factors. The history data with the same or approximative property are used as input of ANN for forecasting. The method of choosing similar day can cover more correlative factors so that we can get more accurate forecasting results.