Graph-Based Semi-supervised Learning for Fault Detection and Classification in Solar Photovoltaic Arrays

Zhao, Yi; Ball, Roy; Mosesian, Jerry; Palma, Jean-François de; Lehman, Brad

doi:10.1109/tpel.2014.2364203

Cited by 269 publications

(84 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The maximum power is calculated by ( for the first string and ( . The total maximum power is (27). Clearly, this maximum power is greater than that for the unarranged arrays (2.6 V module ).…”

Section: A Theoretical Analysismentioning

confidence: 87%

Section: Introductionmentioning

confidence: 99%

“…Paper [26] analyzes the dynamic current-voltage characteristics to achieve fault diagnosis. In paper [27], machine learning techniques are employed for PV fault detection by measuring PV array voltage, current, irradiance and temperature.After a fault is diagnosed, certain remedial measures need to take place. In-situ reconfiguration is an effective solution [28]-[34].…”

mentioning

confidence: 99%

See 2 more Smart Citations

Efficiency Improvement of Nonuniformly Aged PV Arrays

Zhang

et al. 2017

IEEE Trans. Power Electron.

View full text Add to dashboard Cite

Abstract-The utilization of solar energy by photovoltaic (PV) systems have received much research and development (R&D) attention across the globe. In the past decades, a large number of PV array have been installed. Since the installed PV arrays often operate in harsh environments, non-uniform aging can occur and impact adversely on the performance of PV systems, especially in the middle and late periods of their service life. Due to the high cost of replacing aged PV modules by new modules, it is appealing to improve energy efficiency of aged PV systems. For this purpose, this paper presents a PV module reconfiguration strategy to achieve the maximum power generation from non-uniformly aged PV arrays without significant investment. The proposed reconfiguration strategy is based on the cell-unit structure of PV modules, the operating voltage limit of gird-connected converter, and the resulted bucket-effect of the maximum short circuit current. The objectives are to analyze all the potential reorganization options of the PV modules, find the maximum power point and express it in a proposition. This proposition is further developed into a novel implementable algorithm to calculate the maximum power generation and the corresponding reconfiguration of the PV modules. The immediate benefits from this reconfiguration are the increased total power output and maximum power point voltage information for global maximum power point tracking (MPPT). A PV array simulation model is used to illustrate the proposed method under three different cases. Furthermore, an experimental rig is built to verify the effectiveness of the proposed method. The proposed method will open an effective approach for condition-based maintenance of emerging aging PV arrays. Index Terms-Maximum power tracking, non-uniform aging,Manuscript received May 13, 2015; revised October 11, 2015 and January 31, 2016; accepted August 24, 2015. Copyright © 2010 IEEE. Personal use of this material is permitted. However, permission to use this material for any other purposes must be obtained from the IEEE by sending a request to pubs-permissions@ieee.org.Y I. INTRODUCTIONSolar energy utilization has received much attention across the globe over the last decades [1]- [6]. Currently, photovoltaic (PV) power devices are gaining in popularity in the global renewable energy market, primarily owing to the reducing manufacture costs of PV panels and continuous improvement in power conversion technologies [7]- [8]. In practice, high energy conversion efficiency and long effective service time can help reducing capital and operating costs and thus are highly desired.With the improvement of materials technologies, monocrystalline silicon and multicrystalline silicon now can be economically produced in large quantities. However, their energy conversion efficiency from solar to electricity is still low. Typical efficiency for monocrystalline silicon solar cells is around 20% while it is 18% for multicrystalline silicon solar cells [9]. On the power electronics side, hig...

show abstract

Section: A Theoretical Analysismentioning

confidence: 87%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Efficiency Improvement of Nonuniformly Aged PV Arrays

Zhang

et al. 2017

IEEE Trans. Power Electron.

View full text Add to dashboard Cite

show abstract

“…It combines the structure representation and linear discriminant analysis to make a semi-supervised approach, called stable semi-supervised discriminant learning (SSDL). Fault detection for solar photovoltaic cells is proposed based on a novel graph-based semi-supervised learning model, to solve the need of countless labeled data for building a learning model [23]. In [24], an algorithm is introduced to iteratively expand the labeled set by inserting the closest object that was classified as positive to the labeled set.…”

Section: Related Workmentioning

confidence: 99%

Online semi-supervised multi-channel time series classifier based on growing neural gas

Nooralishahi

Seera

Loo

2016

Neural Comput & Applic

View full text Add to dashboard Cite

Challenges in time series classification has attracted attention in the past decade. Although large amounts of labeled data are assumed to be available, in reality, labeled data might be scarce to find in many domains. In this paper, we propose an online semi-supervised multi-channel classifier for time series based on growing neural gas (GNG) learning scheme. The method is able to handle multi-channel time series with variation in dimensions and it introduces a label prediction strategy to minimize misclassification. It measures the similarity of input instance and learned templates using weighted multichannel dynamic time warping technique and learns the topology of input data space specified for each class using the GNG learning algorithm. Comprehensive evaluation is conducted using various datasets, such as gesture recognition, human activity recognition, and human daily-life activity recognition. Experimental results demonstrate good classification results, with indication that the proposed approach requires only a handful of labeled instances to construct an accurate classification model.

show abstract

“…), which reduce both its energetic yield and its efficiency [3]. The central problem with DC side faults is that they are Elyes Garoudja and Kamel Kara are with the Electronic Department, Blida 1 University Blida, Algeria, (g_lyes@hotmail.fr, k.kara68@gmail.com).…”

Section: Introductionmentioning

confidence: 99%