摘要
为准确地预测光伏发电功率,节约资源,提出一种基于改进非线性自回归(nonlinear autoregressive with external input,NARX)神经网络算法的光伏发电功率短期预测模型。通过皮尔森相关分析选择影响发电功率的环境因素,利用遗传算法(GA)优化受限玻耳兹曼机(RBM)模型参数,避免陷入局部最优;利用优化后的RBM模型初始化NARX神经网络的参数。实例预测表明,改进NARX神经网络算法对光伏发电功率短期预测精度更高,收敛速度更快。
In order to accurately predict the amount of photovoltaic power generation and save resources, we proposed a short-term forecast model for photovoltaic power generation(PVPG). This model was based on an improved NARX(nonlinear autoregressive with external input)neural network algorithm. Our study analyzed environmental factors through Pearson correlation analysis. In order to avoid falling into local optimum, our research used genetic algorithm(GA) to optimize the parameters of the restricted Boltzmann machine(RBM). The optimized RBM model used to initialize the parameters of NARX. Experiments show that that our algorithm obviously higher accuracy and the convergence time is shorter than the existing methods.
作者
朱想
丁云龙
郭力
师浩琪
ZHU Xiang;DING Yunlong;GUO Li;SHI Haoqi(Key Laboratory of Smart Grid of Ministry of Education,Tianjin University,Tianjin 300000,China;School of Computer Science,Wuhan University,W uhan 430072,Hubei,China)
出处
《武汉大学学报(理学版)》
CAS
CSCD
北大核心
2020年第5期505-511,共7页
Journal of Wuhan University:Natural Science Edition
基金
国家自然科学基金(51907140)。
关键词
光伏发电
遗传算法
皮尔森相关分析
受限玻耳兹曼机
NARX神经网络
photovoltaic power generation
genetic algorithm
Pearson correlation analysis
restricted Boltzmann machine
NARX(nonlinear autoregressive with external input)neural network