多尺度分析与数据互迁移相结合的短期电力负荷预测方法

Short-term Load Forecasting Method Combining Multi-scale Analysis with Data Co-transfer

为了提升短期电力负荷的预测效果,提出一种多尺度分析与数据互迁移相结合的短期电力负荷预测方法。一方面,针对多尺度分析预测法中分解得到的子序列在建模和预测的过程中没有对原序列中的隐含相关信息加以利用的问题,采用互信息特征选择法选取合适的原负荷序列历史值并将其加入到原负荷序列近似分量的特征集合中,通过特征扩充为学习器提供更多的信息,进而提高近似分量的预测精度。另一方面,针对不同类型的数据在学习过程中会相互影响的问题,采用了一种基于核岭回归的数据互迁移学习方法,将其他几种类型中与待预测类型日相似的数据迁移到待预测类型日的数据中,既利用了这些数据的相似性,又兼顾了这些数据的差异性。测试案例显示,所提方法在MAPE,MAE和RMSE这3个误差评价指标上相对于单模型方法分别降低了6.2%,3.4%和5.5%。

In order to improve the performance of short-term load forecasting,this paper proposed a short-term load forecasting method combining multi-scale analysis with data co-transfer.On the one hand,aiming at the problem that the hidden information in the original series which is related to the subseries isn＇t fully utilized in the modeling and prediction of subseries in multi-scale analysis forecasting method,this paper used mutual information feature selection method to select appropriate past loads and introduced them into the feature set of the approximation component of original load series.By expanding feature set,more information can be provided for learning algorithm,which can further improve the forecasting accuracy of the approximate component.On the other hand,aiming at the problem that different kinds of data can influence model＇s performance,this paper proposed a transfer learning method based on kernel ridge regression to transfer similar data to data corresponding to days to be forecasted.By doing this,the similarity of these data was used and the difference of these data was taken into account during modeling.Case study shows that the proposed method outperforms in MAPE,MAE and RMSE which are decreased by 6.2%,3.4% and 5.5% respectivelywhen compared with single model forecasting method.