基于PSO混合核函数SVM对光伏最大功率点的预测研究

摘要

In view of high cost and low precision of the traditional photovoltaic maximum power point tracking algorithm,this paper presents a scheme for adopting mixed kernel support vector machine algorithm to solve the problem.This paper uses linear weighted Gaussian kernel function and polynomial kernel function to build a mixed support vector machine,and uses particle swarm algorithm to optimize model parameters and predict the voltage at the photovoltaic max.power point.Experimental analysis shows that the mixed kernel function combines the advantages of two single kernels,and considers both the overall and local fitting abilities of the support vector machine.It has a better non-linear function prediction precision than the single kernel model,with strong generalization ability,short training period and good stability.%针对传统的光伏最大功率点跟踪算法存在成本高，精度低的问题，进而采用混合核支持向量机算法尝试解决这一难题。通过研究分析高斯核函数与多项式核函数的特性，线性加权高斯核函数与多项式核函数，构造混合形式的支持向量机模型，并使用粒子群算法优化模型参数，对光伏最大功率点电压进行预测。实验分析表明，混合核函数结合了两个单核的优点，兼顾支持向量机全局拟合能力与局部拟合能力。非线性函数预测精度优于单核模型，泛化能力高，训练时间少，平稳性好。