An effective imaging based diagnosis of

Solar Photovoltaic System and

performance parameters for smart solar

power grid

Kirti Pathak1*, Dr. Sanjay Kumar Jagannath Bagul2

1 Research Scholar, University of Technology, Jaipur, Rajasthan

2 Supervisor, University of Technology, Jaipur, Rajasthan

Abstract - Another of the most attractive power generation sources for residential, corporate,

and commercial operations is energy production. Due to this compelling qualities,

photovoltaic systems (PV) electricity production have increasingly attracted the attention of

researchers and professionals. Nonetheless, the greatest obstacle in generating energy from

the sun is the predictable discontinuous output power from photovoltaic (PV) systems, which

is largely caused by climate. The power differential of a photovoltaic panels may considerably

reduce the financial profit of large scale solar fields. For everyday proper management of

electricity supply production, distribution, and storing as well as for judgement on the

electricity market, results were consistent prediction of the electricity output of solar PV

systems is essential.

The 5 subjects mentioned in this article are:

 The varied faults which might occur in PV panels; internet Photovoltaic panels

monitoring;

 Use of machine learning approaches in Photovoltaic module damage detection;

 Advantages of fault diagnosis in PV panels and the various motion sensors used for

this purpose are discussed.

 Recommendation for future possible research paths are given in view of the evaluated

research.

Keywords - power generation, sources for residential, PV panels, machine learning, electricity

output, large scale solar fields.

INTRODUCTION

Solar power is quickly becoming one of the most attractive alternatives to traditional grid electricity for

residential, business, and industrial use. Solar photovoltaic (PV) systems use PV cells to generate

electricity from sunlight. Photovoltaic (PV) energy, in particular, has gained appeal as a renewable

energy source in recent years because of its low cost, long payback time, and high environmental and

social benefits. One of the most widely used renewable energy sources; solar power is advantageous

due to its low environmental impact, zero operating costs, and unlimited availability. This greatly

accelerated the global rollout of solar photovoltaic (PV) systems.

Inaccurate estimates of solar panels' demise can be caused by environmental factors such as

temperature, cloud cover, dust, irradiance level, and relative humidity. Problems with solar panels can

lead to erratic behaviour in addition to reducing the effectiveness and reliability of the plant's services.

The failure to properly prioritize fault detection can lead to power losses, and when solar arrays are

faulty, the entire system is at risk of failure. In grid-connected solar systems, failure detection using

multiple methods is possible. Some of them use meteorological and astronomical information to spot

problems in GCPV systems.

Use of electricity is now universally recognized as a need. As has been said, rising middle-class and

urban populations are driving up the need for electricity across the world. Given that they are both

non-renewable and polluting, fossil fuels have no place in a sustainable future. Consequently, we

need to reduce our reliance on non-renewable resources, with renewable energy taking on a more

crucial role in the future.

Accurate solar energy projections are critical for photovoltaic (PV) based energy plants to take part in

energy auction markets early and to plan resources efficiently. The literature reports a plethora of

techniques for predicting PV solar output. There are essentially four distinct categories into which

these strategies might be placed:

1. The use of data-driven statistical formulations for the purpose of predicting solar time

series using archival measurements

2. The use of artificial neural networks and other forms of machine learning;

3. Physical models derived from numerical weather prediction and satellite imagery, and

4. Strategies that combine different aspects of the aforementioned approaches are known

as hybrid approaches. The seasonal autoregressive integrated moving average

(SARIMA), a hybrid model combining a random vector functional link neural network and

the discrete wavelet transform, has been introduced as a means of predicting solar PV

power generation over shorter time periods. The findings of the combined models have

been found to be more accurate than those of the separate models.

Pv Systems In Energy Conversion

The efficacy of photovoltaic (PV) systems is heavily dependent on the effectiveness of solar radiation,

temperature, and conversion. PV systems suffer from variable system performance, high installation

costs, and low module efficiency of up to 20% as a direct result of weather fluctuations. This is in

addition to the many other drawbacks that PV systems have. The optimization of the size of stand-

alone or grid-connected photovoltaic systems (GCPVS) is a complex problem of optimization that

predicts the acquiring of adequate energy and economic expenses for customers. With the right

placement of photovoltaic (PV) panels, solar energy can be used practically everywhere, which is one

of its many attractive selling pointsfigure 1.1. In general, the following are some of the benefits of solar

photovoltaic (PV) systems:

 Low environmental effects

 Capable of being conveniently placed near to the customer, hence decreasing the losses that

occur along the transmission line

Figure 1: PV Systems in Energy Conversion

Photovoltaic system design

This section provides information on two different types of job altogether. In the first place, those

works that are associated with the fundamental components of a photovoltaic system, and in the

second place, those works that are associated with the configuration of solar power plants. Figure 1.2

depicts one possible approach to the design of solar systems, which may be found in the

accompanying text.

Figure 2: Design of photovoltaic systems

Photovoltaic system operation

Work related to the general operation of solar systems, the operation of hybrid systems produced by

photovoltaic systems, and the power quality issue will be provided in this section. Photovoltaic

technologies and the problem of power quality combine to generate these hybrid systems. Figure 1.3

depicts one approach to understanding how photovoltaic systems function.

Figure 3: Operation of photovoltaic systems

POWER ELECTRONIC CONVERTERS IN RENEWABLE ENERGY CONVERSION

For the past several decades, scientists have been concentrating their efforts on the study of

renewable energy sources, and now, various power inverters are being manufactured to link these

technologies to the distribution system. In order to guarantee both the delivery of power and the

quality of that power, the transmission lines need to be equipped with high-voltage power electronic

circuits.

Therefore, power electronic inverters are the devices that are tasked with the responsibility of

efficiently accomplishing these conversion operations. The ever-increasing need for energy on a

worldwide scale has resulted in the development of innovative topologies for semiconductors and

power converters that are able to deliver the total amount of power that is necessary. The race to

manufacture semiconductors that are capable of withstanding higher voltage and current for the sake

of more efficient systems is still very much alive and well. In addition, there has been a lot of rivalry

between the use of medium voltage semiconductor devices in current converter topologies and the

use of high voltage semiconductor devices in old converter topologies when it comes to power

conversion.

Traditional Topologies of Multilevel Inverter

There are three primary categories of multilevel inverters, which are denoted as Neutral Point

Clamped (NPC), Flying Capacitor (FC), and Cascaded H-Bridge respectively (CHB).

 Cascaded H-Bridge Multilevel Inverter

In 1975, Baker and Banister invented the CHB inverter, which is recognized as the first MLI to be built

on semiconductors. This topology is distinguished by the presence of a series interface of single-

phase H-bridge inverters. This architecture does not necessitate the use of clamping diodes or

voltage capacitors in any of its circuits. The arrangement in question makes use of independent DC

sources. DC sources can come from a variety of different places, including solar cells, fuel cells, and

batteries. Each individual H-bridge is made up of four switches, each of which has its own diode, in

addition to a separate voltage supply. The architecture of a 9-level CHBMLI is seen here in Figure

2.1.1. The output levels may be boosted by a factor of two simply by adding another individual H-

bridge in cascade. The following equations provide an explanation for the highest voltage levels that

the CHBMLI is capable of producing:

Figure 4: Circuit diagram of 9-level CHBMLI

 Multilevel Inverter with Diode Clamped or Neutral Point Clamped

A Neutral Point Clamped MLI (NPC) was presented for the first time. Figure 2.1.2 depicts a 9-

level NPCMLI arrangement for your viewing pleasure. The NPC is a completely established

arrangement that only uses one DC source. Through the use of capacitors, this DC source is split

up into numerous other DC sources. The NPC is widely implemented throughout all sectors of

industry, often operating at voltages ranging from 2.3 kV to 4.16 kV, with a small number of

applications operating at voltages up to 6 kV. In point of fact, NPC has discovered uses for high-

performance AC drives in the oil and gas sectors, as well as in the mining industry. In addition to

this designation, NPC is also known as Diode Clamped Multilevel inverter (DCMLI).

Figure 5: Neutral or Clamped Point Multilevel Inverter with Clams

STUDY AND PROJECTIONS FOR PV POWER GENERATION

Over the course of the past few decades, a significant number of researches have been devoted to

addressing forecasting issues in a variety of application fields. The application of Recurrent Neural

Networks, often known as RNNs, has proven to be effective in solving machine learning challenges.

These models have been presented as a solution to the challenges associated with time-dependent

learning. The fundamental idea behind RNNs is illustrated in Figure 3; a section of a neural network,

denoted by the letter A, receives an input of the form xt and returns a value denoted by ht. It is

important to highlight that RNNs are well adapted to learn and extract information from temporal

sequences. Given an input sequence x=(x1, x2, xt), the following is a general formula for the RNN

hidden state h t:

Figure 6: Basic illustration of RNN

Character-Level Language Models based on RNN

If you'll think back to our discussion of language modelling, you'll remember that our ultimate goal is to

make predictions about future tokens based on the information we already have about those tokens

and the tokens that came before them (labels). Initially proposed employing a neural network to

represent human language. The following is an example of a language model constructed using

RNNs. Let's assume a minibatch size of 1, and that the text sequence is "machine." We tokenize text

into characters rather than words and examine a character-level language model to facilitate training

in later sections. In character-level language modelling, as shown in Fig. 3.1, an RNN is used to

predict the next character based on the current and prior characters.

Figure 7: RNN-based Character-Level Language Models

REVIEW OF LITERATURE

A multifunctional distributed sparse (DS) control strategy was presented by Singh et al. (2020) for a

solar system. The P&O approach of drawing the most power possible from PV cells is implemented in

the system that has been presented. The recommended system functions as a DSTATCOM in the

event that there is no PV system present.

Solar energy was recommended as the power source for an independent water pumping system by

Mishra and Singh (2020). In the system that is being suggested, the water pumping is done by a

Switched Reluctance Motor (SRM). The P&O method of drawing the maximum amount of electricity

from PV cells is utilized in the water pumping system that has been presented.

A boost DC-DC converter was created by Jhang et al. (2020), and it is built with complementary

metal–oxide–semiconductor (CMOS).

In order to reduce the initial voltage, a logic gate circuit with low leakage has been put into place.

Dehghani et al. (2020) highlighted the benefits that come with utilizing the P&O approach. Both a

P&O-based controller and a fuzzy controller are put through their paces when it comes to simulating

the conditions of a solar PV system that is linked to the grid. When compared to the P&O approach,

the results indicate that the fuzzy controller has a performance that is somewhere in the middle. 0

Mishra and Singh (2020) presented a solar-powered water pumping system that would be

incorporated into an existing grid.

For the solar photovoltaic system, the authors opted for the single step of conversion. A switching

reluctance motor is what gets the job done when it comes to pumping water. The P&O approach is

utilized so that the PV system may provide the greatest amount of electricity that can be harvested.

The P&O approach is the most appropriate choice for tracking maximum power under any

circumstances since photovoltaic (PV) systems are nonlinear by their very nature, and their

performance is mostly determined by the weather conditions.

An improved leader particle swarm optimization (ELPSO) was presented by Ram et al. (2020) for the

P&O approach in order to track the highest power under any unfavourable situations. The improved

leader particle swarm optimization approach is used in order to locate the global maximum power

point (GMPPT). When using the improved P&O approach, one may achieve findings that are both

more accurate and comprehensive.

A drift free P&O approach was presented by Mishra et al. (2020) for the stand alone wind energy

conversion system. The traditional P&O approach generates oscillations and is unable to accurately

track the MPP throughout a broad range of wind speeds. The wind data is being gathered with the

help of the speed sensor, and it will be sent to the suggested P&O technique so that correct power

can be extracted.

THE IMPORTANCE OF MACHINE LEARNING METHODS IN PV PANEL DEFECT DETECTION

Systems trained using machine learning (ML) can analyse data and identify patterns with minimal

human oversight. The production, efficiency, and quality of PV panels are all influenced by their

surroundings and can thus be predicted. Improved forecasting techniques can assist energy providers

and consumers get the most out of these installations. Even while renewables offer lower operating

costs over the long term, the initial investment in equipment is typically rather substantial figure 5. If

the user is able to accurately predict when the grid will be compromised, they will be able to take

preventative measures and save a significant amount of money.

Figure 8: Machine Learning Techniques in Fault Diagnosis

One of the more fruitful uses of machine learning to far has been in the field of predictive analytics.

Machine learning algorithms are applied to the data collected from the equipment's operation to

predict when maintenance is required. By following this procedure, technicians can avoid performing

unnecessary maintenance and costly failures. Human inspectors can try to perform the same tasks

without any assistance, but they typically fail. One study found that predictive maintenance aided by

AI was up to 25.3% more effective and 24.6% more precise than manual maintenance.

An Explanation of the Diverse Sensors Employed in PV Panels for the Detection of Faults

The suggested approach requires both electrical and environmental data to be collected when the PV

panel is in operation. This is necessary in order to achieve the feasible outcomes. The numerous

approaches that have been put into practice by a variety of researchers are the single most important

factor in determining whether or not the PV system is able to successfully detect faults figure 5.1. The

passive component of the diagnostic, also known as fault detection, requires that a normal threshold

and a failure threshold be identified and then fitted.

1.1

2.3

1.2

0.2

1.6

5.1

0.2

0.6 0.5

0.1

0.5

5.1

0.2

0.6

2.3

1.2

0.2

1.6

0.6

1.6 1.4 1.3 1.4 1.2

0.2

1.6

1.2

0.4

1.4 1.2

0.2

1.6

5.1

0.2

0.6

5.1

0.2

1.3

5.1

0.2

0.6

2.3

1.2

0.2

1.6

1.2

0.2

1.6

1.2

0.5 0.5

5.1

0.2

0.6

5.1

0.2

0.6

5.1

2.3

1.2

0.2

1.6

2.3

1.2

1.8

1.4 1.7

1.2

0.2

1.6

1.2 1.2

0 5 10 15 20 25 30 35 40

Current STC Current SC Current Varying

Current Partial shading Current Complete shading Current Degraded

Current Ground fault Current Arc fault

Figure 9: Diverse Sensors Employed in PV Panels

The diagnostics procedure that was based on the model was used to create each residue value that

was displayed here. The non-active component of the diagnostic data was monitored separately with

the immediate power output by the PV system. The fundamental concept behind techniques for flaw

identification is that, for a large solar system, it is necessary to subject the entire string to a wide

variety of irradiance and temperature ranges. It is possible to view the resultant string current, which

is caused by a combination of the types of faults and the locations of the defective modules in the

string.

The Benefits Obtained from the Detection of Defects in PV Panels

Permanent power losses can be classified as defects in solar panels; however, a more fine-grained

analysis may be applicable if there are failure specific patterns that can be exploited. The

identification of the problem leads to an increase in the effectiveness of the solar panels. Performance

is the metric that is used to evaluate the capacity of a solar panel to transform the sun's rays into

usable energy.

At the point when the sun gleams on two sun powered chargers with various evaluations for a similar

timeframe, the board with the higher rating will deliver a more noteworthy measure of energy than the

board with the lower rating. The productivity of sunlight based not entirely settled by how much power

that is delivered by photovoltaic cells.

The proficiency of a sunlight based charger is impacted by the organization of the cells, the electric

game plan of the parts, and different factors. It is crucial to have solar panels with a top-tier efficiency

if you want to get the most out of your energy usage and reduce your bills. Contrasting two sunlight

based chargers, one of which has an effectiveness rating of 21% while different has a productivity

rating of 14%, the board with the higher rating will deliver 50% more kilowatt-hours (kWh) of force

under similar circumstances.

CONCLUSION

This paper compiles and discusses a number of previously published articles that cover recent

developments and research in the following areas: I different potential blames that can happen in PV

board; (ii) on the web/distant oversight of PV boards; (iii) job of AI procedures in shortcoming finding

2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023

Diverse Sensors Employed in PV Panels

Historical data Low scenario High scenario Medium scenario

of PV boards; (iv) different sensors utilized for various issue recognitions in PV boards; and (v)

advantages of shortcoming recognizable proof in PV boards. These points are separated into five

classes: I different According to the viewpoint of issue classification, various other possible

wellsprings of flaws, including halfway concealing issue, hamper, and open circuit shortcoming, as

well as deficiencies in diodes (both hindering and sidestep diodes), were analyzed in more prominent

profundity. In this paper, we examined the different internet based procedures that are intended to

screen the mistakes that happen in PV boards in light of the sort of sensor that is utilized and the

observing of the PV boards. These web-based strategies screen the mistakes that happen in PV

boards in view of the observing of the PV boards. The proposed LSTM-based method to momentary

determining of photovoltaic sun oriented power yield has created results that are very uplifting. Later

on, we intend to carry out and test the exhibition of other RNN models like the Gated repetitive unit

(GRU) model and to consolidate extra data, for example, meteorological information to additionally

work on the exactness of our gauges. This will be important for our endeavours to additional improve

the nature of our figures.

A conversation on the plan, activity, and upkeep of planetary groups has been given. According to an

analysis that was conducted, the most significant breakthroughs in photovoltaic systems are now

being made in the areas of better designs of photovoltaic systems, as well as optimal operation and

maintenance; these are the primary foci of PV systems research.

REFERENCES

1. Antonanzas J, Osorio N, Escobar R, Urraca R, Martinez-de Pison F, Antonanzas-Torres F.

Review of photovoltaic power forecasting. Solar Energy. 2016;136:78-111

2. Behera MK, Majumder I, Nayak N. Solar photovoltaic power forecasting using optimized

modified extreme learning machine technique. Engineering Science and Technology, an

International Journal. 2018;21(3):428-438

3. Chen, Z.; Wu, L.; Cheng, S.; Lin, P.; Wu, Y.; Lin, W. Intelligent fault diagnosis of photovoltaic

arrays based on optimized kernel extreme learning machine and IV characteristics. Appl.

Energy 2017, 204, 912–931.

4. Dhimish, M.; Holmes, V.; Dales, M. Parallel fault detection algorithm for grid-connected

photovoltaic plants. Renew. Energy 2017, 113, 94–111.

5. Duman, S.; Li, J.; Wu, L.; Yorukeren, N. Symbiotic organisms search algorithm-based

security-constrained AC–DC OPF regarding uncertainty of wind, PV and PEV systems. Soft

Comput. 2021, 25, 9389–9426.

6. Eseye AT, Zhang J, Zheng D. Short-term photovoltaic solar power forecasting using a hybrid

wavelet-PSO-SVM model based on SCADA and meteorological information. Renewable

Energy. 2018;118:357-367

7. Harrou F, Dairi A, Taghezouit B, Sun Y. An unsupervised monitoring procedure for detecting

anomalies in photovoltaic systems using a one-class support vector machine. Solar Energy.

2019;179:48-58

8. Harrou F, Taghezouit B, Sun Y. Improved kNN-based monitoring schemes for detecting faults

in PV systems. IEEE Journal of Photovoltaics. 2019;9(3):811-821

9. Harrou F, Taghezouit B, Sun Y. Robust and flexible strategy for fault detection in grid-

connected photovoltaic systems. Energy Conversion and Management. 2019;180:1153-1166

10. Izgi, E.; Öztopal, A.; Yerli, B.; Kaymak, M.K.; ¸Sahin, A.D. Short–mid-term solar power

prediction by using artificial neural networks. Sol. Energy 2012, 86, 725–733.

11. Ji, D.; Zhang, C.; Lv, M.; Ma, Y.; Guan, N. Photovoltaic array fault detection by automatic

reconfiguration. Energies 2017, 10, 699.

12. Qing X, Niu Y. Hourly day-ahead solar irradiance prediction using weather forecasts by

LSTM. Energy. 2018;148:461-468

13. Sobri S, Koohi-Kamali S, Rahim NA. Solar photovoltaic generation forecasting methods: A

review. Energy Conversion and Management.

14. Srivastava S, Lessmann S. A comparative study of LSTM neural networks in forecasting day-

ahead global horizontal irradiance with satellite data. Solar Energy. 2018;162:232-247.

15. Ventura, C.; Tina, G.M. Development of models for on-line diagnostic and energy assessment

analysis of PV power plants: The study case of 1 MW Sicilian PV plant. Energy Proc. 2015,

83, 248–257.

16. Wang H, Yi H, Peng J, Wang G, Liu Y, Jiang H, et al. Deterministic and probabilistic

forecasting of photovoltaic power based on deep convolutional neural network. Energy

Conversion and Management. 2017;153:409-422

17. Wang K, Qi X, Liu H. A comparison of day-ahead photovoltaic power forecasting models

based on deep learning neural network. Applied Energy. 2019;251:113315

18. Wu, J.; Yan, S.; Xie, L. Reliability analysis method of a solar array by using fault tree analysis

and fuzzy reasoning Petri net. Act. Astronaut. 2011, 69, 960–968.

19. Ziar, H.; Manganiello, P.; Isabella, O.; Zeman, M. Photovoltatronics: Intelligent PV-based

devices for energy and information applications. Energy Environ. Sci. 2021, 14, 106–126.