Deep residual network based fault detection and diagnosis of photovoltaic arrays using current-voltage curves and ambient conditions

https://pure.hud.ac.uk/ws/files/20117354/Manuscript.pdf

Artificial neural network based photovoltaic fault detection algorithm integrating two bi-directional input parameters

This paper presents a model-based anomaly detection method for supervising the direct current (dc) side of photovotaic (PV) systems. Toward this end, a framework combining the benefits of $k$ -nearest neighbors ( $k$ NN) with univariate monitoring approaches has been proposed. Specifically, $k$ NN-based Shewhart and exponentially weighted moving average (EWMA) schemes with parametric and nonparametric thresholds have been introduced to suitably detect faults in PV systems. The choice of $k$ NN method to separate normal and abnormal features is motivated by its capacity to handle nonlinear features and do not make assumptions on the underlying data distribution. In addition, because the EWMA approach is sensitive in detecting small changes. First, a simulation model for the inspected PV array is constructed. Then, residuals generated from this model are employed as the input for $k$ NN-based schemes for anomaly detection. Parametric and nonparametric thresholds using kernel density estimation have been used to detect faults. The effectiveness of the $k$ NN-based procedures is verified using actual measurements from a 9.54-kWp grid-connected system in Algeria. Results proclaim the efficiency of the proposed strategy to supervise the dc side of PV systems.

Improved $k$ NN-Based Monitoring Schemes for Detecting Faults in PV Systems

Photovoltaic (PV) systems operating in the outdoor environment are vulnerable to various factors, especially dust impact. Abnormal operations lead to massive power losses, and severe faults as short circuit may cause safety problems and fire hazards. Therefore, monitoring the operation status of PV systems for timely troubleshooting potential failure and effective cleaning scheme are the focus of current research works. In this study, I–V characteristics of PV strings under various fault states are analyzed, especially soiling condition. Because labeled data for PV systems with specific faults are challenging to record, especially in the large-scale ones, a novel algorithm combining artificial bee colony algorithm and semi-supervised extreme learning machine is proposed to handle this problem. The proposed algorithm can diagnose PV faults using a small amount of simulated labeled data and historical unlabeled data, which greatly reduces labor cost and time-consuming. Moreover, the monitoring of dust accumulation can warn power plant owners to clean PV modules in time and increase the power generation benefits. PV systems of 3.51 and 3.9 kWp are used to verify the proposed diagnosis method. Both numerical simulations and experimental results show the accuracy and reliability of the proposed PV diagnostic technology.

Design of Hybrid Artificial Bee Colony Algorithm and Semi-Supervised Extreme Learning Machine for PV Fault Diagnoses by Considering Dust Impact

Fault diagnosis of photovoltaic (PV) arrays is an essential task for improving the reliability and safety of a photovoltaic system (PVS). The PVS faults at the DC side are difficult to detect by traditional protective devices, which may reduce power conversion efficiency and even lead to safety matters and fire disaster. This study investigates a newly-designed fault diagnostic method for a PVS according to the following three steps. First, optimal fault features are extracted by analyzing I-V curves from different faults, including hybrid faults of the PVS under the standard test condition (STC). Moreover, the trust-region-reflective (TRR) deterministic algorithm combined with the particle-swarm-optimization (PSO) metaheuristic algorithm is proposed to standardize fault features into the ones under the STC. In addition, a multi-class adaptive boosting (AdaBoost) algorithm, which is the stage-wise additive modeling using multi-class exponential (SAMME) loss function based on the classification and regression tree (CART) as the weak classifier, is utilized to establish the fault diagnostic model. The effectiveness of the fault diagnostic model could long-term maintain by periodically updating the feature standardization equations to standardize the fault features into the ones under the STC. Various types of the PV modules are used to validate the generalization of the fault diagnostic method. Both the numerical simulations and experimental results show the accuracy and reliability of the proposed fault diagnostic method.

/pdf/newly-designed-fault-diagnostic-method-for-solar-96r5wh50n2.pdf

Newly-Designed Fault Diagnostic Method for Solar Photovoltaic Generation System Based on IV-Curve Measurement

https://hal-centralesupelec.archives-ouvertes.fr/hal-01970593/document

Anne Migan

Papers

PV shading fault detection and classification based on I-V curve using principal component analysis: Application to isolated PV system