PCA-GRNN在综合气象短期负荷预测中的应用

Application of PCA-GRNN in Integrated Meteorological Short-term Load Forecasting

为克服由气象因子较多且信息互嵌造成输入量多、预测时间长、预测精度低的缺点,引入主成分分析(PCA)提取气象因子特征量,与历史负荷数据共同作为建模对象;同时,针对BP神经网络动态性能的不足,建立基于广义回归神经网络(GRNN)的短期负荷预测模型。通过对实际电力负荷数据的预测,证明该方法与传统神经网络预测模型相比,明显提高预测精度和速度,具有实用性和有效性。

In order to avoid the shortcomings such as redundancy inputs, long prediction time and low prediction accuracy, which caused by more weather factors and information embedded each other, principal component analysis （PCA） is adopted to extract characteristics of weather factors which are taken as the modeling objects, together with the dates of historical load. Simultaneously, against the shortcoming of BP neural network under dynamic performance, short-term load forecasting model based on generalized regression neural network （GRNN） is established. Comparing with the traditional network model, the forecast to the actual power system load proved that, this method can improve the prediction accuracy and speed significantly and was more practical and effective.