基于改进原子轨道搜索算法优化随机森林分类器的光伏系统故障诊断

PV system fault diagnosis based on random forest classifier optimized by improved atomic orbital search algorithm

针对光伏系统故障难以被准确高效地诊断和分类的问题,提出了一种基于改进原子轨道搜索优化的随机森林(IAOS-RF)算法。此算法在光子的发射和吸收部分引入了自适应权重机制和反向学习机制,用于更新电子的位置,能有效加强算法在搜索空间的全面勘探和开发能力。基于一组并网光伏系统故障数据进行算例分析,对比了IAOS-RF算法和几类基准算法的性能差异,结果显示,IAOS-RF算法故障分类准确率最高并且可达到98%,其诊断结果趋于稳定所需的迭代次数最小,具有较快的收敛速度。最后针对该算法存在的一些局限性和改进空间,提出未来需要进一步研究和探讨的问题。

The random forest classifier optimized by improved atomic orbital search algorithm(IAOS-RF)is applied in PV system fault diagnosis and classification to improve the accuracy and effective.This algorithm introduces adaptive weight mechanism and reverse learning mechanism to photon emission and photon absorption to update the position of electrons,which can effectively enhance the algorithm's comprehensive exploration and development capabilities in the search space.Based on a set of fault data from a grid-connected PV system,the differences in performances between the proposed improved algorithm and fundamental algorithms are compared.The results showed that IAOS-RF has the highest fault classification accuracy among the algorithms,reaching 98%.At the same time,its diagnosis,with a fast convergence rate,requires the least times of iterations to be stable.In the end,in the view of the limitations in the proposed algorithm,the problems need to be improved in the future are discussed.