基于RRPNN-CEEMD-BiLSTM的短期电力负荷预测  

Short-Term Power Load Forecasting Based on RRPNN-CEEMD-BiLSTM Model

作　　者：魏晓宾 焦丕华 胡钰业 于洋 WEI Xiao-bin;JIAO Pi-hua;HU Yu-ye;YU Yang(School of Electrical and Electronic Engineering,Shandong University of Technology,Zibo Shandong 255000,China;Shandong Deyou Electric Co.Ltd.,Zibo Shandong 255000,China)

机构地区：[1]山东理工大学电气与电子工程学院,山东淄博255000 [2]山东德佑电气股份有限公司,山东淄博255000

出　　处：《计算机仿真》2024年第11期137-141,共5页Computer Simulation

基　　金：山东省重点研发计划项目资助(2019JZZY020804)。

摘　　要：随着低碳经济的不断发展,准确的电力负荷预测对于能源管理至关重要。为了提高电力负荷短期预测的准确度,提出基于改进的岭多项式神经网络(RRPNN)、互补集合经验模态分解(CEEMD)和双向长短期记忆网络(BiLSTM)的短期电力负荷预测。首先通过自回归输入调整RRPNN的输出精度,实现快速处理非线性负荷数据并实现初始预测。然后,通过CEEMD和BiLSTM方法,减少模态混叠和模态间的相互影响,从而获得准确的模态分量预测结果。最后,以某地区实际电网负荷为例,进行仿真验证。通过验证可知,与RRPNN-BiLSTM、CEEMD-BiLSTM、RRPNN-PSO-BiLSTM等其它模型相比,RRPNN-CEEMD-BiLSTM模型能够有效实现短期电力负荷的精确预测,具备较高的负荷预测精度。With the continuous development of low-carbon economy,accurate power load forecasting is crucial for energy management.To improve the accuracy of short-term power load forecasting,short-term power load forecasting based on an improved ridge polynomial neural network(RRPNN),complementary ensemble empirical modal decom-position(CEEMD)and bi-directional long short-term memory network(BiLSTM)is proposed.Firstly,the output ac-curacy of the RRPNN is adjusted by autoregressive inputs to enable fast processing of non-linear load data and to a-chieve initial forecasts.Then,the CEEMD and BiLSTM methods are used to reduce modal confounding and inter-mo-dal interactions to obtain accurate modal component forecasting results.Finally,a simulation test is carried out to veri-fy the actual grid load in a region as an example.The experimental validation shows that the RRPNN-CEEMD-BiL-STM model can effectively achieve accurate short-term power load forecasting with high load forecasting accuracy compared with other models such as RRPNN-BiLSTM,CEEMD-BiLSTM and RRPNN-PSO-BiLSTM.

关 键 词：短期负荷预测 岭多项式神经网络 互补集合经验模态分解 双向长短期记忆网络 

分 类 号：TP743[自动化与计算机技术—检测技术与自动化装置]