The poor outdoor operating conditions of household photovoltaic (PV) make the power station prone to various faults. However

the dispersion of household PV installations often increases the difficulty and cost of operation and maintenance (O&M). Although the remote monitoring and fault detection of a PV power station can be realized by the use of operation data

the particularity of a household power station also brings many problems to fault detection. In this study

we propose a shading fault detection method of household PV power based on inherent characteristics of monthly string current data mapping. The ideal current peak obtained by a new fitting method is used to normalize string current data. The current probability density function (PDF) at each time point is estimated by kernel density estimation (KDE). Through the normalized current data corresponding to the maximum probability density

the inherent characteristics of the strings are obtained

such that whether the strings have shading can be judged and the shading degree can then be evaluated. Not only are no additional sensors needed to collect environmental data

such as irradiation and temperature

but also simulating the detailed parameters of the power station is not required. The interference caused by meteorological factors can thus be eliminated

which can be easily used in old power stations and newly constructed power stations. The effectiveness and performance of the proposed shading fault detection method is verified by experimental data collected from the actual household PV power station.