Search

Aiming at improving the effectiveness and real-time performance of the fault detection of the photovoltaic (PV) array under different outdoor conditions, a novel method based on Grubbs criterion and local outlier factor (G-LOF) method is proposed in this study. The simulation model of the PV array is built to obtain the reference current of each PV string. The local outlier factor (LOF) is a ratio of the average local reachability density of sample neighbourhoods relative to its local reachability density. In this study, the LOF is used to quantify the abnormality by comparing the measured current and the simulated reference current. To reduce the false alarm probability of LOF, the Grubbs criterion is utilised as the test criterion to detect the abnormality of PV strings and it is merged with the LOF to derivate the G-LOF. The experiments of partial shading, open-circuit fault and short-circuit fault of PV modules are implemented in sunny and cloudy days. Experimental results verify that the proposed G-LOF method can sensitively detect above abnormalities, especially can detect the slight performance reduction caused by the partial shading or faults.

Aiming at evaluating the state of the photovoltaic (PV) array and improving the reliability of the PV system, a fault diagnosis method for PV arrays based on fault parameters identification is proposed in this paper. Compared with existing fault diagnosis methods, the proposed method has advantages of recognizing concurrent faults and describing each fault quantitatively by identifying fault parameters from the measured current-voltage (I–V) curve of the PV array. The methodology consists of three parts. Firstly, functional relationships between unknown parameters in the one-diode model of PV cells with environmental parameters are obtained by parameters extraction. Secondly, a code-based fast fault simulation model (CFFSM) is established to simulate I–V curves of the PV array under various faulted conditions. Thirdly, by determining the fault parameters to be identified and constructing an objective function that is the error between the simulated I–V curve with the measured I–V curve, an optimization problem is formulated, in which optimal fault parameters are identified by applying the differential evolution (DE) algorithm. The validation experiments in summer and early spring show that the proposed diagnosis method can identify the parameters of up to three concurrent faults, including partial shading, short circuit, and increased series-resistance losses, under good irradiance condition with high accuracy.

The large scale application of the photovoltaic (PV) systems is significantly beneficial to the mitigation of energy crisis. The quality and performance of PV systems directly influence the energy yield. It is necessary to establish more scientific and effective methods to maximize the energy yield. In this paper, the concept of health status is proposed to describe the performance of PV systems within a certain period of time. A health status based performance evaluation model is built by the Gaussian mixture models (GMM) and the empirical mode decomposition (EMD). Then, the health index (HI) of PV array is defined. Under the outdoor ambient conditions, the proposed model can sensitively detect the slight performance reduction caused by faults or partial shadings of the PV array in real-time. At last, the proposed method is verified by the simulations and experiments. Experimental results show that on the sunny day the average daily HI is 1.2 and the average performance ratio (PR) is 0.85, which both show the PV array is healthy. When one of the PV modules in the PV array is partially shaded, the PR is still approximate 0.9. However, the calculated HI is greater than the threshold and the fault is reported. The results indicate that the proposed method can sensitively identify pseudo health status that cannot be identified by the PR. The proposed evaluation method based on health status provides an alternative option to assess the performance of the PV systems. Combining with the PR, the comprehensive performance of the PV systems can be reflected more accurately.

Aiming at monitoring photovoltaic (PV) systems, evaluating the degree of fault and locating the fault automatically under different outdoor conditions, this paper discusses a new procedure of fault detection and evaluation of fault degree of the PV system. For the PV array connected by PV modules in series and parallel, each string shares the same voltage. The value of current can be used to identify the underperformed strings. In addition, considering the non-stationary stochastic characteristics of current of PV strings, the local outlier factor (LOF) is applied to detect the fault in PV system by evaluating the deviation between the observed data and the whole data. Nevertheless, the LOF method is more suitable for large samples and the LOF value varies with the value of string current. Hence, the conventional LOF method is not suitable for evaluating the fault degree. In order to apply this method to different scale PV systems to detect the fault accurately and evaluate the fault degree, a modified algorithm is proposed in this study. The simulations and experiments based on the model of PV array in MATLAB/Simulink and the 10 kWp PV power plant built on the campus of Hohai University are implemented. The results of experiments reveal that the modified LOF has good performance in fault detection and fault degree evaluation in different scales of the PV systems.