基于改进回归法的电力负荷预测

Power Load Forecasting Based on Improved Regression

提出了一种基于改进回归法的电力负荷预测方法,在对历史数据进行分析和参数估计的基础上,先用岭回归法剔除奇异值,再用主成分回归法提取影响负荷的主要因素,得出模型的解析形式。针对实际系统的应用验证了该方法不仅适用于短期负荷预测,也适用于超短期负荷预测。此外建立了一些特定因素的模糊函数,在超短期负荷预测过程中采用了聚类分析法提取负荷相似日。通过不同的简化,该方法可蜕化为传统的Kalman预测、相似日预测和神经网络预测,是一种比较全面的负荷预测方法,可得出高精度的预测结果。

A load forecasting method based on improved regression is proposed. On the basis of historical data analysis and parameter estimation, firstly the singular values are rejected by ridge regression, then the main factors impacting the load arc extracted by principal component regression and the analytical form of the model is obtained. Applying this method to actual power system, the results show that not only the proposed method is suitable to short-term load forecasting, but also to ultra short-term load forecasting. In addition, the fuzzy functions for specified factors are built, and in ultra short-term load forecasting the similar load days are extracted by clustering analysis. Through different simplification the proposed method can be transferred into traditional Kalman filter forecasting, similar day forecasting and neural network forecasting respectively, therefore this method is a relatively comprehensive load forecasting method and by this method the forecasting results with high accuracy can be obtained.