基于脊波递归神经网络的短期负荷预测模型
|
摘要
为准确有效地预测电力系统负荷,本文提出了一种基于脊波递归神经网络(ridgelet recurrent neural network,RRNN)的电力系统短期负荷预测模型。采用脊波函数作为预测模型隐含层神经元中的激励函数,在网络模型的内部加入关联层,关联层节点可有效存储隐含层神经元的内部状态,增强了网络模型的反馈联接,并采用粒子群优化算法(particle swarm optimization,PSO)对网络的参数和权值进行优化,避免模型出现陷入局部最优解的问题。以某地区电网负荷系统作为仿真算例,对预测模型进行仿真和测试。仿真结果表明,与BP神经网络和常规脊波神经网络预测模型相比,本文所提出的预测模型的日平均绝对误差和日最大绝对误差均有所降低,具有较好的预测性能。该研究具有一定的实际应用价值。
In order to predict power system load accurately and effectively,a short-term load forecasting model of power system based on the ridgelet recurrent neural network(RRNN)is proposed in this paper.The RRNN used the ridge wave function as the excitation function in the hidden layer neuron of the neural network.At the same time,the associated layer was introduced in the network model.The nodes in the association layer could effectively store the internal states of the neurons in the hidden layer and enhanced the feedback connection of the network model.Particle swarm optimization(PSO)is used to optimize the network parameters and weights to avoid the model getting into the local optimal solution.The regional grid load system was used as a simulation example to simulate and test the prediction model.The simulation results show that compared with the BP neural network and the conventional ridge neural network prediction model,the average absolute error and the maximum absolute error of the prediction model proposed in this paper are both reduced,and the predictive model has better predictive performance.This study has a certain practical value.
In order to predict power system load accurately and effectively,a short-term load forecasting model of power system based on the ridgelet recurrent neural network(RRNN)is proposed in this paper.The RRNN used the ridge wave function as the excitation function in the hidden layer neuron of the neural network.At the same time,the associated layer was introduced in the network model.The nodes in the association layer could effectively store the internal states of the neurons in the hidden layer and enhanced the feedback connection of the network model.Particle swarm optimization(PSO)is used to optimize the network parameters and weights to avoid the model getting into the local optimal solution.The regional grid load system was used as a simulation example to simulate and test the prediction model.The simulation results show that compared with the BP neural network and the conventional ridge neural network prediction model,the average absolute error and the maximum absolute error of the prediction model proposed in this paper are both reduced,and the predictive model has better predictive performance.This study has a certain practical value.
引文
[1]王惠中,周佳,刘轲.电力系统短期负荷预测方法研究综述[J].电气自动化,2015,37(1):1-3.
[2]廖旎焕,胡智宏,马莹莹,等.电力系统短期负荷预测方法综述[J].电力系统保护与控制,2011,39(1):29.
[3]乔维德.基于改进PSO-BP神经网络的短期电力负荷预测研究[J].电力系统保护与控制,2007,35(17):17-21.
[4]张平,潘学萍,薛文超.基于小波分解模糊灰色聚类和BP神经网络的短期负荷预测[J].电力自动化设备,2012,32(11):121-125.
[5]Ulagammai M,Venkatesh P,Kannan P S,et al.Application of Bacterial Foraging Technique Trained Artificial and Wavelet Neural Networks in Load Forecasting[J].Neurocomputing,2007,70(16):2659-2667.
[6]贺晓,刘爱国,孙蕾,等.基于小波神经网络的中长期电力负荷预测[J].陕西电力,2012,40(2):4-8.
[7]曹双华,曹家枞,李涛,等.基于小波变换的神经网络空调负荷预测研究[J].暖通空调,2005,35(4):13-17.
[8]陈标.短期电力负荷的小波神经网络预测[D].长沙:湖南大学,2012.
[9]张智晟,段晓燕,李伟婕,等.优化动态递归小波神经网络短期负荷预测模型[J].电力系统及其自动化学报,2009,21(5):30-35.
[10]徐建军,杜蔷楠,闫丽梅.基于对角递归神经网络的小区用电负荷预测研究[J].化工自动化及仪表,2017,44(11):1052-1055.
[11]刘丽,王曙钊,李敬社.脊波变换的研究现状与应用前景[J].太赫兹科学与电子信息学报,2005,3(4):315-319.
[12]茆美琴,周松林,苏建徽.基于脊波神经网络的短期风电功率预测[J].电力系统自动化,2011,35(7):70-74.
[13]于惠鸣,撖奥洋,于立涛,等.基于PSO-DNN的电力系统短期负荷预测模型研究[J].青岛大学学报:工程技术版,2017,32(2):62-66.
[14]王倩,张智晟,王帅,等.基于PSO-RNN的光伏发电功率预测研究[J].青岛大学学报:工程技术版,2014,29(4):40-43.
[15]Candes E J.Ridgelet:Theoryand Applications[D].Stanford:Department of Statistics Stanford University,1998.
[16]Figueiredo E M N,Ludermir T B,Figueiredo E M N,et al.Investigating the Use of Alternative Topologies on Performance of the PSO-ELM[J].Neurocomputing,2014,127(3):4-12.
[17]周任军,李绍金,李红英,等.空间粒子群优化算法及其在电力系统环保经济负荷分配中的应用[J].电力自动化设备,2014,34(9):7-12.
[18]陈鸿琳,李欣然,冷华,等.运用PSO和GRNN的短期负荷二维组合预测[J].电力系统及其自动化学报,2018,30(2):85-89.
[19]陈贵敏,贾建援,韩琪.粒子群优化算法的惯性权值递减策略研究[J].西安交通大学学报,2006,40(1):53-56.
[20]陆宁,周建中,何耀耀.粒子群优化的神经网络模型在短期负荷预测中的应用[J].电力系统保护与控制,2010,38(12):65-68.
[2]廖旎焕,胡智宏,马莹莹,等.电力系统短期负荷预测方法综述[J].电力系统保护与控制,2011,39(1):29.
[3]乔维德.基于改进PSO-BP神经网络的短期电力负荷预测研究[J].电力系统保护与控制,2007,35(17):17-21.
[4]张平,潘学萍,薛文超.基于小波分解模糊灰色聚类和BP神经网络的短期负荷预测[J].电力自动化设备,2012,32(11):121-125.
[5]Ulagammai M,Venkatesh P,Kannan P S,et al.Application of Bacterial Foraging Technique Trained Artificial and Wavelet Neural Networks in Load Forecasting[J].Neurocomputing,2007,70(16):2659-2667.
[6]贺晓,刘爱国,孙蕾,等.基于小波神经网络的中长期电力负荷预测[J].陕西电力,2012,40(2):4-8.
[7]曹双华,曹家枞,李涛,等.基于小波变换的神经网络空调负荷预测研究[J].暖通空调,2005,35(4):13-17.
[8]陈标.短期电力负荷的小波神经网络预测[D].长沙:湖南大学,2012.
[9]张智晟,段晓燕,李伟婕,等.优化动态递归小波神经网络短期负荷预测模型[J].电力系统及其自动化学报,2009,21(5):30-35.
[10]徐建军,杜蔷楠,闫丽梅.基于对角递归神经网络的小区用电负荷预测研究[J].化工自动化及仪表,2017,44(11):1052-1055.
[11]刘丽,王曙钊,李敬社.脊波变换的研究现状与应用前景[J].太赫兹科学与电子信息学报,2005,3(4):315-319.
[12]茆美琴,周松林,苏建徽.基于脊波神经网络的短期风电功率预测[J].电力系统自动化,2011,35(7):70-74.
[13]于惠鸣,撖奥洋,于立涛,等.基于PSO-DNN的电力系统短期负荷预测模型研究[J].青岛大学学报:工程技术版,2017,32(2):62-66.
[14]王倩,张智晟,王帅,等.基于PSO-RNN的光伏发电功率预测研究[J].青岛大学学报:工程技术版,2014,29(4):40-43.
[15]Candes E J.Ridgelet:Theoryand Applications[D].Stanford:Department of Statistics Stanford University,1998.
[16]Figueiredo E M N,Ludermir T B,Figueiredo E M N,et al.Investigating the Use of Alternative Topologies on Performance of the PSO-ELM[J].Neurocomputing,2014,127(3):4-12.
[17]周任军,李绍金,李红英,等.空间粒子群优化算法及其在电力系统环保经济负荷分配中的应用[J].电力自动化设备,2014,34(9):7-12.
[18]陈鸿琳,李欣然,冷华,等.运用PSO和GRNN的短期负荷二维组合预测[J].电力系统及其自动化学报,2018,30(2):85-89.
[19]陈贵敏,贾建援,韩琪.粒子群优化算法的惯性权值递减策略研究[J].西安交通大学学报,2006,40(1):53-56.
[20]陆宁,周建中,何耀耀.粒子群优化的神经网络模型在短期负荷预测中的应用[J].电力系统保护与控制,2010,38(12):65-68.