基于优化VMD和组合模型的电力负荷预测  

Power Load Forecasting Based on Optimized VMD and Combined Model

作　　者：倪自聪 黄德启[1] NI Zi-cong;HUANG De-qi(College of Electrical Engineering,Xinjiang University,Urumqi Xinjiang 830017,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830017

出　　处：《计算机仿真》2025年第3期128-133,共6页Computer Simulation

基　　金：新疆维吾尔自治区重大科技专项(2022A01007)。

摘　　要：电力负荷预测在电力网络的规划、运行和决策中具有极为重要的功能。为了提高负荷预测的准确性和鲁棒性,提出了一种基于卷积神经网络、双向长短期记忆网络和极限梯度提升的电力负荷预测方法。首先采用白鹭群优化算法变分模态分解的核心参数进行优化,使用优化后的变分模态分解对电力负荷数据进行分解,将原始信号分解为多个时频局部化模态函数,以有效提取信号的时频特征。其次结合卷积神经网络和双向长短期记忆网络,实现对复杂非线性的电力负荷序列模式的识别和建模,并结合极限梯度提升构建一个用于电力负荷预测的组合模型。通过实际算例对所建立的电力负荷预测模型进行仿真分析,验证了所提出的组合预测模型能有效地提高电力负荷预测的精确度和稳定性。Power load forecasting has an extremely important function in the planning,operation and decision making of power networks.In order to improve the accuracy and robustness of load prediction,a power load prediction method based on convolutional neural network(CNN),bidirectional long and short-term memory(BiLSTM),and extreme gradient boosting(XGBoost)is proposed.Firstly,the core parameters of the variational modal decomposition of the Egret swarm optimization algorithm(ESOA)are optimized,and the optimized variational modal decomposition(VMD)is used to decompose the electric load data,and the original signal is decomposed into multiple time-frequency localized modal functions to effectively extract the time-frequency characteristics of the signal.Secondly,convolutional neural network and bidirectional long and short-term memory are combined to realize the recognition and modeling of complex nonlinear power load sequence patterns,and a combined model for power load prediction is constructed by combining extreme gradient boosting.Through practical calculation examples,the established power load forecasting model is simulated and analyzed.It is verified that the proposed combined forecasting model can effectively improve the accuracy and stability of power load forecasting.

关 键 词：电力负荷预测 卷积神经网络 双向长短期记忆 极限梯度提升 组合模型 

分 类 号：TM743[电气工程—电力系统及自动化]