该文主要针对光伏系统直流侧常见且危害较大的串联电弧故障问题展开研究。首先，搭建了光伏系统电弧故障实验平台，并利用该平台采集了正常状态及电弧故障状态下工作电流的数据，建立了试验数据库。随后，利用快速傅里叶变换对故障电弧的频率成分进行分析，确定了故障电弧的特征频带，并利用该特征频带选定了适用于串联电弧故障检测的最优小波函数。通过分析故障电弧信号最大值及小波细节系数d1的方差、模极大值，提出了3种能有效检测出电弧故障状态的时频域判据，并发现不同负载电流、不同电压及不同功率对这3种故障判据并无影响，最后提出一种基于时域与频域的电弧故障混合判据，该判据较单一判据具有更高的可靠性，为光伏系统直流电弧检测奠定了理论基础。“,”This paper mainly studied the seriously harmful series arc fault of photovoltaic systems. First of all, PV system arc-fault test platform was established to collect normal and fault working current, and a test database was built. After that, fast Fourier transformation (FFT) was employed to determine the characteristic band of arc-fault, and the optimal wavelet function was chosen based on the characteristic band. By analyzing the maximal value of the arc fault signal and the variance and modulus maxima of the wavelet detail coefficients d1, this paper proposes three time and frequency domain criterions to detect the arc fault and finds out that different load current, voltage and power have no effect on the three criterions. Finally, a new mixed criterion is presented, which has higher reliability and lays an important foundation for the photovoltaic system DC arc detection.