基于深度学习集成方法的日用电最大负荷预测

Daily electric peak load forecasting based on SDAE-B

电力负荷受气温等多重因素影响,具有短期波动性和非线性特征,给电力系统调度带来了极大不确定性和挑战。为提前做好电力生产计划和调度预案,开展短期电力日最大负荷预测具有重要的应用价值。基于深度学习集成(SDAE-B)方法的计算准确度和计算效率优势,对较大变化的外界因素具有高鲁棒性。选取某省级电网2018—2020年三年的日度电力数据和气温数据,利用SDAE-B方法对该地区2020年任意15天日最大负荷进行预测,并与运用SDAE方法和支持向量回归(Support Vector Regression,SVR)方法得到的测试结果进行比较。结果显示SDAE-B方法的预测误差最小,且该深度学习方法具有强大的特征提取能力,能在最大程度上减少数据典型特征的损失,且很好地跟踪电力日最大负荷的非线性特征。

The electric load in power market is of significant randomness and non-linearity,which brings great uncertainty and challenge to power system scheduling.In order to make power dispatching plan,the electric load peak forecast have a wide potential application in power areas.The Stacked Denoising Auto Encoder--Bagging(SDAE-B)method has the advantages of calculation accuracy and efficiency,and high robustness to large changes in external factors.In this paper,the daily power grid unified adjustment data and temperature data of a Province in China from 2018 to 2020 are selected to predict the maximum load of any 15-days dispatching in 2020 in the Province using SDAE-B method,and the test results are compared with those obtained by SDAE method and SVR method.The results show that the deep learning integration method(SDAE-B)has the minimum error in the prediction,so the SDAE-B method is used to forecast the short-term power load peak with less error,and a relatively satisfactory result is achieved in the prediction accuracy.The empirical results show that the proposed deep learning method has strong feature extraction ability and robustness,which can minimize the loss of typical features of data and track the nonlinear characteristics of electric daily peak load.