基于改进BP-SVM-ELM组合预测的光伏MPPT方法研究
|
摘要
针对传统最大功率点跟踪方法存在功率振荡和跟踪速度慢的问题,文章提出一种考虑影响光伏输出特性因素变量的组合预测方法。该方法使用遗传算法优化逆向传播神经网络、最小二乘支持向量机和极限学习机分别预测最大功率点对应的电压,然后再通过方差-协方差权值动态分配法来组合预测。通过仿真实验分析,验证了该组合预测方法能够利用各算法自身的优势,并有效地避免其不足,从根本上提高了预测模型的性能。通过与传统的扰动观测法对比,确认采用该方法不仅能保证光伏阵列能够稳定运行在最大功率点,而且有效地缩短了跟踪最大功率点的时间,提高了光伏发电系统效率。
In order to solve the problems of maximum power point tracking(MPPT) such as power oscillation and slow tracking speed when using traditional methods, this paper proposed a combined prediction algorithm that considered variables affecting photovoltaic output characteristics. The algorithm uses genetic algorithm to optimize BP neural network, least squares support vector machine and extreme learning machine(ELM) to predict the voltage of maximum power point respectively, and then adopts variance-covariance(VC)weight dynamic allocation method to combine the predictions. Through experimental simulation analysis, the combined forecasting method can use the advantages of each algorithm, effectively avoiding their deficiencies, and fundamentally improve the performance of the predictive model. Compared with the traditional disturbance observation method, the combined prediction algorithm not only ensures the stable operation of photovoltaic array at the maximum power point, but also effectively saves the time for tracking the maximum power point, which is of great significance to improve the efficiency of photovoltaic power generation system.
In order to solve the problems of maximum power point tracking(MPPT) such as power oscillation and slow tracking speed when using traditional methods, this paper proposed a combined prediction algorithm that considered variables affecting photovoltaic output characteristics. The algorithm uses genetic algorithm to optimize BP neural network, least squares support vector machine and extreme learning machine(ELM) to predict the voltage of maximum power point respectively, and then adopts variance-covariance(VC)weight dynamic allocation method to combine the predictions. Through experimental simulation analysis, the combined forecasting method can use the advantages of each algorithm, effectively avoiding their deficiencies, and fundamentally improve the performance of the predictive model. Compared with the traditional disturbance observation method, the combined prediction algorithm not only ensures the stable operation of photovoltaic array at the maximum power point, but also effectively saves the time for tracking the maximum power point, which is of great significance to improve the efficiency of photovoltaic power generation system.
引文
[1]德勤研究.2017清洁能源行业报告[J].电器工业,2017(11):6-26.
[2]张兴,曹仁贤.太阳能光伏并网发电及其逆变控制[M].北京:机械工业出版社,2011.
[3]范钦民,闫飞,张翠芳,等.基于模糊控制的光伏MPPT算法改进[J].太阳能学报,2017,38(8):2151-2158.
[4]黄鹏.基于神经网络预测模型的光伏发电系统MPPT研究[D].北京:华北电力大学,2016.
[5]张智慧.基于支持向量回归的光伏发电MPPT算法研究[D].杭州:浙江大学,2013.
[6]王薇.混沌改进群体智能算法研究及其在光伏MPPT中的应用[D].南昌:南昌大学,2017.
[7]PAPPU S M J,GUMMADI S N.Artificial neural network and regression coupled genetic algorithm to optimize parameters for enhanced xylitol production by Debaryomyces nepalensis in bioreactor[J].Biochemical Engineering Journal,2017,120(1):136-145.
[8]朱霄珣,韩中合.基于PSO参数优化的LS-SVM风速预测方法研究[J].中国电机工程学报,2016,36(23):6337-6342,6598.
[9]何延昭,王贞艳,郑世强.基于在线最小二乘支持向量机逆系统的高速永磁同步电机解耦控制[J].中国电机工程学报,2016,36(20):5639-5646,5741.
[10]刘彩霞,方建军,刘艳霞,等.基于多类特征融合的极限学习在四足机器人野外地形识别中的应用[J].电子测量与仪器学报,2018,32(2):97-105.
[11]吕智林,谭颖,李捷,等.基于Markov-ELM的独立混合微网分布式电源多目标容量优化配置[J].中国电机工程学报,2017,37(7):1927-1936.
[12]YADAV B,CH S,MATHUR S S,et al.Assessing the suitability of extreme learning machines(ELM)for groundwater level prediction[J].Journal of Water and Land Development,2017,32(1):103-112.
[2]张兴,曹仁贤.太阳能光伏并网发电及其逆变控制[M].北京:机械工业出版社,2011.
[3]范钦民,闫飞,张翠芳,等.基于模糊控制的光伏MPPT算法改进[J].太阳能学报,2017,38(8):2151-2158.
[4]黄鹏.基于神经网络预测模型的光伏发电系统MPPT研究[D].北京:华北电力大学,2016.
[5]张智慧.基于支持向量回归的光伏发电MPPT算法研究[D].杭州:浙江大学,2013.
[6]王薇.混沌改进群体智能算法研究及其在光伏MPPT中的应用[D].南昌:南昌大学,2017.
[7]PAPPU S M J,GUMMADI S N.Artificial neural network and regression coupled genetic algorithm to optimize parameters for enhanced xylitol production by Debaryomyces nepalensis in bioreactor[J].Biochemical Engineering Journal,2017,120(1):136-145.
[8]朱霄珣,韩中合.基于PSO参数优化的LS-SVM风速预测方法研究[J].中国电机工程学报,2016,36(23):6337-6342,6598.
[9]何延昭,王贞艳,郑世强.基于在线最小二乘支持向量机逆系统的高速永磁同步电机解耦控制[J].中国电机工程学报,2016,36(20):5639-5646,5741.
[10]刘彩霞,方建军,刘艳霞,等.基于多类特征融合的极限学习在四足机器人野外地形识别中的应用[J].电子测量与仪器学报,2018,32(2):97-105.
[11]吕智林,谭颖,李捷,等.基于Markov-ELM的独立混合微网分布式电源多目标容量优化配置[J].中国电机工程学报,2017,37(7):1927-1936.
[12]YADAV B,CH S,MATHUR S S,et al.Assessing the suitability of extreme learning machines(ELM)for groundwater level prediction[J].Journal of Water and Land Development,2017,32(1):103-112.