IoT-based green-smart photovoltaic system under extreme climatic conditions for sustainable energy development

0 Introduction

Sustainable development based on green energy has become a universal objective of humankind worldwide.As a green energy source,solar energy offers numerous advantages such as environmental friendliness,costeffectiveness,and lower operational and maintenance costs [1].However,photovoltaic (PV) systems face numerous challenges in extreme conditions.On Earth,these conditions include strong wind-sand environment,intense ultraviolet (UV) radiation,heavy salt mist environment,and extremely cold weather,which can result in a series of faults such as hot spots,bubbles,and short circuits,compromising the long-term stability and performance of PV systems,as shown in Fig.1.Space environments also present challenges such as extreme temperatures on the lunar surface,lunar dust,and cosmic radiation [2-4].The role of sensing devices is becoming increasingly vital for ensuring the reliable functioning of PV systems [5].However,given the above conditions,sensing technologies face a number of challenges in terms of the measurement accuracy,data fluctuations,and sensor materials.This calls for the development of advanced sensing technologies and materials to meet these requirements.

Fig.1 Numerous faults of PV modules in extreme weather conditions: (a) hot spots caused by the accumulation of sand and dust in strong wind-sand environment,(b) bubbles caused by the disintegration of cross-linking agent in strong sunlight conditions,(c) conductive channels caused by the accumulation of sodium ions in heavy salt mist environment,eventually resulting in potential induced degradation (PID),and (d) short circuit caused by water droplets on the icicles entering the junction box in extremely cold weather

Fig.2 Pathway of IoT-based green-smart photovoltaic system under extreme climate

In the era of digital transformation,the importance of utilizing Internet of Things (IoT)-based digital solutions to enhance quality,reduce costs,and conserve energy is being increasingly recognized [6].With steady advances in intelligent automation and digitalization,smart energy systems have emerged concurrently in the Internet of Everything (IoE) era,which is based on air-land-sea-space integrated information network [7].A multifunctional monitoring system based on various sensors in smart energy systems can efficiently monitor the environment and provide preventive warnings under extreme climatic conditions.The system enables operators to perform preventive maintenance,thereby mitigating the impact of these conditions on the efficiency of PV modules [8].For example,an off-grid PV system located in the Sahara region,designed to power small greenhouse farms,employs an IoTbased monitoring system to remotely track parameters such as solar irradiance,PV voltage,and cell temperature [9].This system can accurately detect and diagnose faults in real time.

Therefore,it is crucial to integrate advanced IoTbased sensing technologies into PV systems.Traditional monitoring techniques primarily rely on conventional devices such as current and voltage sensors.However,the reliability and accuracy of these sensors are often compromised under extreme climatic conditions [10,11].In contrast,the emerging sensor technologies reviewed in this paper offer several advantages,including higher measurement precision and stable performance in harsh environments characterized by extreme temperature,humidity,and intense UV radiation [12].Advanced sensors such as fiber optic sensors for temperature measurement and wireless humidity sensors are specifically designed to withstand environmental stressors that typically degrade traditional monitoring devices.Furthermore,these advanced sensors are well suited for distributed measurement systems,enabling more detailed real-time data collection across various sections of the PV system [13].By processing sensor data,functionalities such as fault detection,efficiency optimization,and predictive capabilities can be effectively achieved [14,15].

However,despite the continuous advancements in IoT-based technologies in PV systems,there are several limitations with these technologies,which require attention.For instance,efficiently processing large volumes of realtime data and addressing the degradation of sensor materials,which can lead to inaccurate measurements under extreme climatic conditions,are among the main challenges of IoTbased technologies [16].In addition,the data reliability is undermined by network latency or signal loss during remote monitoring under harsh weather conditions.With recent developments,these challenges are being gradually addressed.Advances in sensor materials have improved device durability under extreme environmental conditions[17,18],and improvements in data-processing algorithms have enabled more accurate and efficient analysis of largescale real-time data [19].

In view of the above context,we present a comprehensive review in this paper,which focuses on sensor related topics in IoT-based green-smart PV systems under harsh weather conditions.This review focuses on sensing technologies,data-processing methodologies,and intelligent sensing materials.Figure 2 highlights the structure of this paper:In Section 2,we introduce the climatic parameters and sensing technologies for extreme weather conditions.In Section 3,we describe the comprehensive processing of sensor data.In Section 4,we summarize the smart materials and their related aging phenomena.Finally,we present the conclusions in Section 5.

1 Review screening methods

We conducted this review based on content analysis.The Web of Science,IEEE Xplore,ScienceDirect,and Google Scholar databases were searched using a combination of multiple keywords or phrases,such as “photovoltaic,”“extreme climate,” “high temperature,” “sensors,”“prediction,” “smart materials,” and “power generation efficiency.” The primary sources for this search were peerreviewed English-language publications published between 2013 and 2024.

The titles and abstracts were initially screened to exclude studies that did not focus on green-smart PV systems or extreme climatic conditions.Subsequently,the full texts of potentially relevant articles were assessed against the following inclusion criteria: (1) emphasis on green-smart PV systems,(2) investigation of the performance or design of PV systems under at least one extreme climatic condition,and (3) modeling or simulation studies with clear validation.

The methodologies of the studies reviewed in this paper primarily consisted of literature review,survey methods,empirical analysis,experimental methods,qualitative analysis,quantitative analysis,and observational methods.In this review,we primarily adopted a literature-based approach to evaluate various strategies for enhancing the efficiency and durability of PV systems in extreme climatic environments.

2 Parameter monitoring and sensing technologies

Sensors are crucial to monitor parameters and detect the performance of PV systems in extreme weather conditions,which not only facilitate operators to assess the operating status of PV panels,but also promotes the rational and efficient use of solar energy [20].The IEC 61724-1:2021 standard specifies the requirements for sensor monitoring of various environmental factors,including temperature,humidity,wind speed,and wind direction.According to this standard,temperature sensors should record both ambient air and PV module temperatures with an uncertainty of ±1 °C and resolution of ≤ 0.1 °C or better.For wind speeds,the measurement uncertainty of the sensor must be 0.5 m/s at a speed of 5 m/s and 10% of the reading for speeds exceeding 5 m/s.In this section,we present the sensors used to monitor the environmental parameters of PV systems under harsh weather conditions.The performance and features of each sensor are summarized in Table 1.

Table 1 Sensors for monitoring extreme environmental conditions

2.1 Strong wind-sand environment

In Morocco,the wind speeds during extreme wind events exceed 16 m/s [21].In the eastern Hexi Corridor,the peak wind velocity is 38 m/s [22].Hence,dust and particulate matter (PM) may accumulate on the surface of PV panels when PV systems operate in gusty wind-sand environments[23].These sediments reduce the transmittance of the PV panels.An experimental study demonstrated that as the density of sand increased on the surface of the PV modules,the output power of the PV modules significantly decreased.Specifically,when the sand density increased from 0 g/m²to 40 g/m²,the output power decreased by as much as 32.2% [24].The accumulation of dust on PV panels can reduce the PV output power by up to 50% in Qatar [25].In desert regions,power losses can reach 80%,significantly affecting the overall performance of solar energy systems[26,27].In addition,the degradation in performance is not only due to the physical blocking of sunlight,but also the erosion caused by sand particles.One study found that at erosion velocities of 25 and 30 m/s,the output power of PV modules decreased by 9.82%-16% and 15.42%-24.46%,respectively,compared with that of uneroded modules for various angles of erosion [28].Another study found that,owing to the erosive effects of dust accumulation and other factors,the optimal lifetime of PV panels in Saharan regions is limited to 10 years [29].Thus,monitoring the wind speed,wind direction,and PM concentration are crucial for PV systems in gusty wind-sand environments.

2.1.1 Sensors for detecting wind speed and direction

There are two categories of wind speed and direction sensors,each having different functional elements.Li et al.presented an arc ultrasonic sensor based on array signal processing algorithms for measuring the wind velocity and direction [30].This sensor can accurately determine the wind direction across a 360° range with a precision of 1°,and can gauge wind speeds from 0 m/s to 60 m/s with a detailed accuracy of 0.1 m/s.To reduce the measurement inaccuracies for both wind speed and direction,an ultrasonic sensor array with a semi-conical design was proposed,which can concurrently assess wind parameters in a threedimensional space [31].The results demonstrated that the errors in the wind speed and direction converged toward zero when the signal-to-noise ratio was over 15 dB.

Furthermore,a system based on self-powered triboelectric sensors was used to gauge the wind velocity and direction.Wang et al.[32] developed a sensor based on a triboelectric nanogenerator (TENG).The sensor consisted of a wind vane(v-TENG) and an anemometer (a-TENG).The experiments demonstrated that the a-TENG can detect wind velocities ranging from 2.7 m/s to 8.0 m/s,whereas the v-TENG can effectively monitor eight distinct wind directions.

The FST200-204 sensor is a commercial sensor launched by Firstrate Sensor that demonstrates superior performance.This sensor employs the ultrasonic time difference technique to accurately determine the wind velocity and direction,and has great potential for monitoring the wind parameters of PV systems [33].This sensor can measure wind speeds from 0 m/s to 60 m/s with a sensitivity and accuracy of 0.1 m/s and ±0.3 m/s,respectively.This sensor can also measure wind direction from 0° to 360°,achieving a sensitivity of 0.1° and an accuracy of ±2°.

2.1.2 Sensors for detecting PM concentration

Two types of sensors with different functional components are used for monitoring the concentration of PM in the atmosphere.The first type is PM sensor based on silicon microfabrication technology,which offers advantages in terms of compact size and affordability[34].The results indicated the sensor can attain a precision below 10 μg/m3.The advantages of this sensor include its facile integration and rapid response.The second type is a microsensor based on surface acoustic wave (SAW)resonator [35].The experiments revealed that the SAW sensor can detect masses below 1 ng,showing a remarkable mass sensitivity of 275 Hz/ng.The DL0001 laser dust sensor is a commercial sensor developed by AUDIOWELL,which demonstrates outstanding performance in measuring airborne PM.The DL0001 laser dust sensor is capable of monitoring PM concentrations within a range of 0-500 μg/m3 and diameter of 0.3-10 μm,which is promising for IoTbased PV systems [36].

PM in gusty wind-sand environments significantly influences the stability and efficiency of PV systems [51].PM sensors assist in monitoring the effects of PM on the performance of the PV modules [52].At the same time,using PM sensors,the operators can obtain information on the PM concentration in the air in a timely manner,and formulate cleaning plans to ensure that the PV panels remain in optimal conditions [53].Thus,PM sensors are not only an important tool for monitoring the performance of PV systems,but they are also a key component for protecting PV systems from wind and sand [54].

2.2 Intense UV radiation

In the south of 41°N in Europe,the UV index (UVI) can reach a maximum of 12.7 [55].The peak UVI often exceeds 20 in the Atacama Desert in Chile [56].Under intense UV radiation,the protective coating of PV systems may detach,exposing the PV modules to UV radiation [57,58].Long-term exposure causes discoloration and degradation of the surfaces of PV modules [59],thereby reducing the performance of PV systems [60,61].To minimize the adverse effects of UV radiation on the efficiency and lifetime of PV systems,UV sensors are employed to monitor the UVI [62].

Among the materials suitable for UV sensors,zinc oxide(ZnO) nanostructures are highly promising because of their outstanding electrical,optical,and energy characteristics,as well as ease of preparation [63].As a response characteristic,the on/off ratio is a crucial metric for UV sensors [64].Liao et al.[37] developed a UV sensor based on highly stretchable ZnO.The sensor exhibited a significant on/off ratio of 158.2,along with various physical characteristics,making it highly suitable for multiparametric sensing platforms such as in devices with wearable technology.The response and recovery times of UV sensors are also crucial under intense UV radiation [38].Caliendo et al.[65] investigated a UV sensor based on SAW in ZnO/fused silica,whose response and recovery times were 10 and 13 s,respectively.Moreover,owing to the advantages of SAW sensors in terms of wireless information transmission,UV sensors based on ZnO/fused silica are suitable candidates for UV detection in extreme weather conditions.

Gallium nitride (GaN) and silicon carbide (SiC) are considered ideal materials for manufacturing commercial UV sensors because they overcome the limitations of silicon-based sensors,which are highly sensitive to visible radiation yet demonstrate reduced sensitivity to UV radiation.The GS-ABC-2835M UV sensor introduced by GaNo Opto is characterized by its compact size,high sensitivity,resistance to visible light interference,low power consumption,and high reliability [39].This sensor offers a spectral response range of 210-370 nm.This sensor is typically used to detect UV radiation and it has great potential to monitor the UVI of PV systems.

2.3 Heavy salt mist environment

PV systems in coastal areas,such as those in the central California coast,typically face severe corrosion issues owing to the presence of high salt concentrations [66,67].When the PV system operates in a heavy salt mist environment,chloride and other salt particles are deposited on the surface of the PV panels and other components [68].During partial discharge (PD) activity indirectly caused by salt mist,the double bonds of polyethylene terephthalate (PET),which is a common material for manufacturing backsheets,are attacked,resulting in backsheet degradation [69].These salt deposits can damage the PV modules,leading to electrical failures and corrosion [70],which in turn,result in short circuits,earthing problems,arc discharge,and fire hazards[71].One study showed that using salt mist sensors to monitor the salt mist concentration can aid in understanding the deposition status of salt in a timely manner and avoid breaking the interfacial bonds between the encapsulant and adjacent layers [72].Therefore,real-time monitoring of salt mist concentration using sensors is vital to avoid such incidents [73].

2.3.1 Sensors for detecting salt mist concentration

A salinity sensor based on optical fiber using a signal interferometer (SI) is an important sensor for detecting salt mist concentrations.To expand the scale of salinity measurements,Aslam et al.[40] presented a salinity sensor based on optical SI with ultrahigh sensitivity.The sensitivity of the sensor could reach up to 7.5 nm/% within a salinity range of 0%-100%.Quantifying salt mist concentrations via conductivity measurements is also a practical technology[74].In 2021,a reusable sensor based on electrical conductivity (EC) was investigated [41].The validation results revealed that the average errors in the laboratory and field measurements were 6 and 11%,respectively.

Among the commercial salinity sensors,the Y521-A four-electrode conductivity (salinity) sensor launched by Yosemitech outperforms traditional two-electrode designs by offering enhanced accuracy,stability,and a broader range of measurements.This model is unaffected by polarization and requires less maintenance over long periods,making it promising for IoT-based green-smart PV systems.The conductivity of this sensor spans from 0.01 Ms/cm to 5 Ms/cm or up to 100 Ms/cm,maintaining an accuracy below 1% or precisely 0.01 Ms/cm.Its salinity detection extends from 0 ppt to 2.5 ppt (or even up to 80 ppt),with an accuracy of±0.05 ppt or ±1 ppt,respectively [42].

2.3.2 Sensors for detecting RH

Atmospheric humidity is a crucial factor in the formation of salt mist [75,76].High RH facilitates the accumulation of salt particles and sticky dust layers on PV surfaces,potentially resulting in deposition of salt particles and reducing the power output of PV systems [77].It was observed in a study that the PV power generation was reduced by 40% during the rainy season when the RH reached 76.3%,and by 45%under overcast conditions when the RH was 60.45% [78].RH sensors not only aid in measuring the RH levels and provide real-time data but also offer early warning to operators,enabling them to swiftly implement measures to minimize losses in power output [79,80].

A capacitive humidity sensor based on polyimide was proposed in 2018 [43].The measured capacitance was linearly correlated with the RH level when the RH was within a range of 5%-85%.The RH sensor was shown to have rapid response and recovery times.In 2019,an optical humidity sensor based on nanocomposite film was developed [44].This sensor was expensive and the experimental results showed that this sensor had high stability.In 2020,an RH sensor was developed based on the sensing properties of organometallic perovskite CH3NH3PbI3 [45].This sensor had remarkable sensitivity for detecting RH because its core components decomposed faster in water [81].Hence,this sensor has great potential for monitoring real-time changes in the atmospheric humidity of PV systems.

The SHT45-AD1F sensor,launched by Sensirion,is a SHT4x series industry-proven commercial digital RH sensor.The SHT45-AD1F sensor operates within a range of 0-100% RH,with an accuracy of ±1.0% RH and a humidity response time of 4 s [46].Its long-term stability and high precision ensure reliable measurements.Moreover,this sensor has enhanced dust and water protection,and it is designed to be robust and durable,making it well suited for monitoring PV systems in extreme weather conditions.

2.4 Extremely cold weather

At the Amundsen-Scott South Pole station,there was a continuous period of 78 days in which the maximum temperatures remained at or below -50 °C [82].When a PV system operates in extremely cold weather,the PV modules are covered with ice and snow [83],which hinders the absorption of incident light and results in a decline in the efficiency and electricity production of the PV panels[84].Therefore,icing sensors are crucial for predicting the occurrence of freezing events of PV systems at extremely cold temperatures [85].

An optical load sensor was developed to monitor icing,and the experiments demonstrated that its resolution and sensitivity were 20.4 N and 0.04903 pm/N,respectively[47].In contrast to the traditional strain-gauge load cell,this sensor offers several advantages,including strong resistance to electromagnetic interference,passivity,and a long lifetime.In 2020,an icing estimation technology based on distributed fiber optic sensors was proposed [48].This method can provide temperature information,identify the starting time of icing,estimate the quantity of icing,and ascertain the correlation between the variation in icing and temperature.In 2021,an icing sensor based on phase discrimination was developed that used a three-element electrostatic array to acquire signals with linear decoupling properties [49].

The JCF-1610 icing sensor introduced by SENSOR JC is commercial icing sensor that employs microwave disturbance technology,which makes it suitable for direct ice monitoring in PV systems.This sensor is compact,lightweight,and has precise ice detection capabilities with a resolution of up to 0.1 mm.This sensor operates within a temperature range of -40-85 °C,and can detect ice thickness from 0.1 mm to 4 mm [50].

In summary,the challenges associated with the reliability and performance of PV systems need to be addressed [86].However,real-time monitoring of various data through sensors can facilitate operators in predicting the potential problems of PV systems in a timely manner and optimize their cleaning and maintenance plans for PV systems [87,88].Thus,it is necessary to deploy sensors to monitor PV systems,particularly under various extreme weather conditions,as shown in Fig.3.

Fig.3 Intelligent monitoring platform for photovoltaic systems: (a) monitoring wind speed,direction,and dust particles in strong wind-sand climate,(b) monitoring solar irradiance and UV intensity in intense UV,(c) monitoring salt mist concentration and humidity in heavy salt mist,(d) monitoring the icing sensor current in extreme cold weather

3 Data science process

With the rapid development of IoT,sensor data have emerged as a predominant source of data collection in various domains,such as PV systems,environmental monitoring,and healthcare [89].A large amount of data is collected from sensors during operation of the PV system.By implementing advanced techniques,such as machine learning (ML) and data mining,valuable information can be extracted from this vast repository of sensor data [90].The data processed through preprocessing and data mining can be supplied to algorithms in order to realize various functionalities.In addition,security evaluation,health status evaluation,and maintenance of PV systems can be conducted based on data acquired from smart sensors in IoT-based green-smart PV systems,and maintenance standards can be developed.We present the algorithms and models in detail in this section.

3.1 Data preprocessing

As the penetration of PV generation advances by leaps and bounds,power systems are easily influenced by PV generation.Moreover,the performance of a grid-connected system is influenced by the local weather at a specific location,and the stochastic characteristics of harsh weather severely influence the PV power generation [91].Owing to data communication malfunctions,technical failures,and electromagnetic interference under extreme weather conditions,the PV system encounters some implausible values such as missing and outlier data,which affect the reliability of the entire dataset and cause inaccurate or misleading decision-making results [92].Nevertheless,this can be alleviated by data preprocessing,which can improve the quality of the raw data [93].Here,we discuss the methodologies for improving the quality of the compromised sensor data.

Data cleaning is effective in enhancing the quality and consistency of data by identifying and eliminating erroneous,missing,and inconsistent information [94].Given the abundance of outliers during PV system operation,data cleaning is routinely performed for data preprocessing of PV systems [95].Researchers have proposed various algorithms to clean incorrect data.Among them,Li et al.[96] and Wang et al.[97] employed quartile and clustering algorithms,which enabled the analysis of how anomalous data were distributed along the irradiance-power curve,facilitating the recognition and mitigation of abnormal data.Compared with clustering algorithms,Jiao et al.[95]proposed a method based on sparse dataset.Their method was based on identifying the features of variability within the PV power data and their strong associative relationships with the irradiance data,thereby offering effective analysis.This method can be used for preprocessing data.In addition,this method demonstrates versatility in extending cleansing of anomalous data of other time series.

Data imputation is also a prevalent technique for enhancing the quality of datasets for PV systems and has been shown to be effective for improving the performance of PV modules [98,99].Koubli et al.[99] developed a method for replenishing missing meteorological and electrical data during the operation of PV systems.Using this method,the augmented data can be used to estimate the output of PV systems.Although the literature on handling missing PV data is limited,potential methods can be derived from other industries,such as medicine [100,101] and biology [102-104].By migrating one variation of generative adversarial network (GAN) used for medical data,Zhang et al.[105] proposed the solarGAN method,which is an innovative approach aimed at the imputation of multivariate solar data that can address relatively independent solar timeseries data.Compared with other GAN-based methods,the error of the solarGAN method decreased by no less than 23.9%.Furthermore,in terms of handling the challenges associated with data imputation in PV power generation series,the neural network model architecture demonstrated exceptional performance,particularly when dealing with missing values ranging from 30% to 70%,and most notably at 50% [106].

In addition to the aforementioned methods,Livera et al.[107] proposed a method for PV data processing,quality validation,and data reconstruction.This method encompasses preliminary statistical evaluation,consistency checks,filtering,detection,and handling of invalid or missing data,and dataset aggregation.Furthermore,this method can provide high-quality,refined performance data for data-driven monitoring systems.

Given the impact of harsh weather conditions,the amount of data that can be captured by sensors may be constrained.To address this issue,data augmentation techniques have been proven to be practical and effective.Researchers have widely explored these techniques for PV systems and utilized them to realize various applications such as detecting series arc faults [108] and expanding datasets of electroluminescence (EL) images [109].

3.2 Data-powered functionalities

Under extreme weather conditions,the operational status,power generation efficiency,and remaining lifetime of PV systems are severely threatened.For instance,various malfunctions occur during extreme ice-rich weather,including insulator flashovers and breakages of lines coated with ice [110].With a large amount of data captured by the sensors,these situations can be alleviated using algorithms that realize the following functionalities: fault detection and diagnosis (FDD),maximum power point tracking (MPPT),lifetime prediction,and load prediction.This is shown in Fig.4.By employing diverse algorithms,the stability and efficiency of PV systems can be guaranteed,and at the same time,these algorithms are beneficial for optimizing maintenance planning and load prediction.

Fig.4 Application scenarios of algorithms used in PV systems

3.2.1 Fault detection

FDD techniques for PV systems are gaining importance to ensure productivity and safety of the systems [111].Hence,several fault detection algorithms for PV modules and inverters that can effectively identify various types of faults have been developed.

PV modules may become susceptible to corrosion [112],potential induced degradation (PID) [113],and electrical issues under extreme weather conditions.Papari et al.[114] achieved an accuracy of 95.4% for fault detection by adopting algorithms based on adaptive neuro-fuzzy inference system (ANFIS),and Abbas et al.[115] achieved a remarkable accuracy of 99.9% by employing subtractive clustering and grid partitioning techniques in conjunction with ANFIS. In contrast to the above algorithms,imagebased algorithms are capable of offering the states of PV modules,including their temperature distributions,hot spots,and other relevant features [116].Using algorithms based on the analysis of infrared thermography (IRT)images,Alajmi et al.[117] achieved an accuracy of up to 100% in three specific scenarios: (1) a healthy system,(2) a system with a transient hot spot issue,and (3) a system with a permanent hot spot issue.Another alternative technique for detecting faults involves using algorithms to analyze the I-V curve,which can sensitively detect faults in the absence of meteorological data [118].

The algorithms for detecting inverter faults can be categorized into two groups based on the voltage or current signals.By analyzing the time waveforms of the output voltage of the inverter,the algorithm can detect an insulated gate bipolar transistor (IGBT) open circuit with an average accuracy of 95.81% under 13 environmental conditions[119].In addition,another algorithm was developed that could detect a series arc within 30 ms by analyzing the PV current noise signal,which could differentiate inverter and arc noise through the analysis of their periodic traits,and the discrepancy of the fluctuation features was then employed to identify the presence of a series arc [120].

3.2.2 Enhancement of power generation efficiency

Because PV systems rely heavily on environmental conditions,the output power of PV systems is uncertain[121].When PV panels face partial shading conditions(PSCs),there are multiple maximum power points (MPPs),which make it difficult to track the global maximum power point (GMPP) [122].To enable PV arrays to achieve optimal performance,it is essential to implement the MPPT algorithm [123,124].To further enhance the power output under PSCs,PV reconfiguration techniques can be employed.By rearranging the physical or electrical connections of the PV modules,the impact of shading can be mitigated,and the power output can be improved [125].

Some researchers have proposed metaheuristicbased MPPT algorithms,offering an efficient and elegant approach for handling various issues such as high computational complexity and dependence on training data [126].Refaat et al.[127] proposed a MPPT method based on an Enhanced Autonomous Group Particle Swarm Optimization (EAGPSO) algorithm,exhibiting a MPPT efficiency of 98.6% and achieving rapid tracking in just 2.6 s.Nevertheless,metaheuristic algorithms present several drawbacks such as challenging parameter tuning,difficulties in distinguishing PSCs,and inability to ascertain the causes of power output variations [128].To address these issues,Mohammed et al.[128] proposed an improved Rat Swarm Optimizer algorithm (IRSO),which could successfully resolve the above issues and prevent unnecessary areas from being searched,thereby achieving faster tracking in less than 1 s,with an average efficiency of 99.89%.

Given the large and rapid variations in irradiance under partially extreme weather conditions,some MPPT algorithms may fail to respond promptly and are unable to capture the optimal working point and accurately track the MPP [129,130].Therefore,improved MPPT algorithms are required to address sudden irradiance changes.To address this issue,two hybrid MPPT algorithms were developed,which exhibited robust performance in recovering from sudden changes in weather conditions and coping with PSCs,with an efficiency of over 97% [131].Moreover,an algorithm combining artificial neural network (ANN) with a traditional MPPT algorithm was developed,which could operate even in the absence of an irradiance sensor and achieved a 15.3% increase in energy output under transient shading conditions [132].

Although these techniques were designed to locate the GMPP,they did not effectively address the issue of multiple power peaks [133].To overcome this challenge,researchers proposed a novel solution,which is reconfiguring the PV array using an array reconfiguration technique [134].The core idea behind this approach is to rearrange the PV panels to distribute shading uniformly across the array,thereby maintaining a stable current output [135].Building on this concept,several methods have been developed to enable the efficient reconfiguration of PV panels.

There are two primary methods used to reconfigure PV arrays: dynamic and static techniques [136].Pillai et al.[137] proposed a column index-based static reconfiguration method for PV arrays,whereas Venkateswari et al.[138]proposed a static reconfiguration technique based on Lo Shu.Both methods effectively mitigate the impact of shading from any location.Compared with static techniques,dynamic techniques are more flexible because they allow real-time adjustment of electrical connections between modules in response to changes in irradiance [139].Aljafari et al.[140] proposed a dynamic reconfiguration method based on dragonfly optimization,which offered the advantages of high reliability,low complexity,and rapid operation.Furthermore,Alharbi et al.[141] proposed a PV array reconfiguration method based on the war strategy optimization algorithm,which enhanced power generation under various shading conditions by minimizing the absolute difference between the maximum and minimum currents across the rows of the array.

3.2.3 PV parameter identification

Given the high initial investment costs associated with solar power utility grids,the development of proper PV system models,particularly for PV modules,is crucial for optimizing the system design and precisely assessing the performance [142].However,owing to the nonlinear characteristics of PV power generation and its sensitivity to external conditions such as temperature,light intensity,and load characteristics,there is a need to improve methods for estimating the model parameters [143].

Consequently,this area of research has garnered considerable attention in recent years.Kumar et al.[144]introduced a modified Rao-based dichotomy technique to identify the actual parameters of various PV cells,which was particularly well suited for rapidly changing irradiation conditions and demonstrated strong adaptability.Furthermore,Ebrahimi et al.[145] introduced a Flexible Particle Swarm Optimization (FPSO) algorithm to estimate the parameters of PV cell models,aiming to enhance the performance of traditional particle swarm optimization techniques.Similarly,Nunes et al.[146] proposed a multiswarm spiral leader particle swarm optimization(M-SLPSO) metaheuristic algorithm for parameter identification of PV models.This approach provides a highly reliable and accurate solution for parameter estimation even under unstable operating conditions.

3.2.4 Lifetime and load prediction

The lifetime of PV modules may be influenced by a range of stress elements during operation,including solar radiation,elevated temperatures,and moisture [147].In addition,distributed PV systems have experienced considerable expansion in the past few years,highlighting the challenge of accurately forecasting net loads [148,149].Given these issues,we present the methods for lifetime and load forecasting in this section.

The service lifetime prediction of PV modules facilitates in establishing reasonable warranty terms and enhances the longevity of PV components [150].Rizzo et al.[151]proposed an algorithm capable of predicting the lifetime of devices and adapting to various shapes of aging curves.Aitio et al.[152] proposed a method based on ML that could achieve an accuracy of 73% in predicting the end of battery life eight weeks ahead.Several methods have been developed based on models for predicting the lifetime of PV modules.Using the gamma process model,the lifetime at failure and the condition variations at inspection can be accurately expressed as probability distributions[153].Kaaya et al.[154] proposed a hybrid approach that integrated the strengths of both the physical model and datadriven algorithms,thereby enhancing prediction accuracy.

In addition,accurate load forecasting facilitates rational arrangement of power production,transmission,and distribution [155].Researchers have proposed load forecasting models based on ML.Wen et al.[156] proposed a model based on a deep recurrent neural network with long short-term memory units (DRNN-LSTM).This model can predict short-term residential power loads and overcome the transient and nonlinear characteristics of the residential electricity load curves.Furthermore,Jurado et al.[157] proposed an improved encoder-decoder-based convolutional neural network (CNN) model that leveraged the combination of deep learning techniques and possessed the capability to forecast short-term net loads and considered the effects of meteorological and temporal variables.

3.2.5 Security and health status evaluation

During the operation of PV systems,their security may be compromised by internal and external factors,resulting in large fluctuations and randomness in the output power[158],as shown in Fig.5.By introducing the health status,the detailed performance of PV systems can be described and quantified [159].Hence,we present the various methods for evaluating the security and health status of PV systems in this section.

Fig.5 Inspections based on appearance defects

Security means that a power system can operate correctly even during unforeseen circumstances [160].Researchers have developed methods to evaluate the security of power systems.A real-time method for evaluating and classifying static security,namely C4.5,was proposed,which can effectively evaluate and classify static security with correct classification rates of～97.44%and～97.74% and computational time of 0 s in the training phase [161].In addition,Dhandhia et al.[162] used support vector machine (SVM) to handle the classification issue of power system security.The states of the PV systems were categorized as secure,alarm,and insecure by an SVM classifier,with a classification accuracy of 98.5% for the IEEE 118-bus test system.

There are several methods for evaluating the health status of PV systems,each of which uses distinct data types.Ding et al.[159] proposed an evaluation model that adopted voltage and current as the characteristic parameters of health status.Compared with the performance ratio (PR),which is an indicator used to measure the efficiency of PV systems [163],this model outperformed PR under PSCs and obtained the same health status as PR on sunny days.In addition,Ding et al.[164] developed another evaluation approach that used I-V characteristics,which could identify the PV array that required fault analysis and showed better sensitivity than direct current PR in detecting the health condition of PV arrays.Moreover,Oviedo et al.[165]proposed a method based on the MPP current to predict three states of PV panels: health,big snail trail,and broken glass.

3.3 Related standards

During the construction and maintenance of PV systems,it is imperative to strictly comply with the relevant standards,which are essential not only for preventing risks such as electric shocks,fires,and injuries due to mechanical and environmental stresses,but also to ensure the stability,long-term reliability,and efficiency of the PV system.

The partial requirements for safe construction of PV systems according to category are presented in Table 2.

Table 2 Requirements for safe construction of PV systems

Furthermore,systematic examination and maintenance of components are vital to guarantee the consistent performance of PV systems.Based on the IEC 62446-2 standard,component inspections and related maintenance are elaborated.As shown in Table 3,the inverter and PV modules are used as examples to elaborate the partial requirements.

Table 3 Requirements for maintenance of PV systems

However,it is worth noting that the above maintenance measures rely to a great extent on manual inspection,which is easy to implement,but is time-consuming and inefficient[166].This demonstrates that there is still a lack of passive and high-precision sensor technologies that can significantly enhance the speed and accuracy of maintenance and improve the reliability of PV systems.

In addition to existing standards,the Middle East and North Africa (MENA) region are anticipated to play an increasingly critical role in global carbon reduction,and it is projected that this region will supply 40% of the world’s energy by 2050 [167].Consequently,PV system testing in this region will become increasingly pivotal,and it is likely that current standards will evolve to address the unique conditions and requirements of this region.

Special geographic conditions,such as in Morocco,North Africa,are also important to develop PV system testing procedures,maintenance standards,and commercial markets.For instance,Morocco is a desert region,which experiences extreme weather conditions such as dust storms[168],high temperatures [169],and aridity,which makes Morocco one of the best candidates for PV system empirical testing.International cooperation with Morocco for the optimization of PV system standards can be a promising pathway for the development of sustainable energy.

4 Smart materials and their aging phenomena

The methods for data quality enhancement,fault detection,and performance improvement to mitigate the impact of extreme climate conditions on sensor data quality,reduce the frequency of failures of PV systems,and enhance system performance have been discussed above.In this section,we present the promising smart materials and the aging phenomenon of sensors and PV panel materials to further enhance the stability and performance of IoT-based green-smart PV systems,as shown in Fig.6.Smart materials are gaining prominence owing to their unique properties[170].By leveraging smart materials such as piezoelectric[171-173],electrocaloric [174-176],and ferroelectric [177,178] materials,significant advantages can be gained in enhancing sensor sensitivity,storing monitoring data,and boosting the power generation efficiency of PV panels.This section aims to elucidate the development of innovative sensor materials and encourage researchers to delve deeper into the development of sensors suitable for extreme climatic conditions.

Fig.6 The schematic of smart materials and insulating materials

4.1 Smart materials

In order to optimize the design of sensors and PV panels under extreme weather conditions,we discuss several promising smart materials,which possess the potential to enhance the performance of sensors and PV panels.

Piezoelectric smart materials are used to convert mechanical energy into electricity [171].This property makes piezoelectric materials particularly suitable for use in environments subject to mechanical stresses,such as winds or seismic activity,where they can generate supplementary power and enhance the overall system efficiency [179].They can be used to fabricate sensors that monitor the PV components to detect and prevent potential failures and damage,thereby improving the stability and safety of PV systems [180].There are many piezoelectric materials with excellent piezoelectric coefficients (d33),such as K0.5Na0.5NbO3-based ceramics (d33=490-570 pC/N)[173] and BaTiO3-based materials (d33=～620 pC/N) [172],which have great potential for use as sensitive components of the sensors as well as used to monitor the status of pressure [181],electric field distortion [182],and magnetic field detection [183].Piezoelectric materials are ideal for converting mechanical stress such as wind or seismic activity into electrical energy,making them effective in areas prone to high winds or frequent tectonic movements.

Furthermore,the data collected by the sensors are stored in memory for subsequent operations such as data analysis and prediction maintenance.The integrated system based on the computing in memory (CIM) unit and sensing may be a promising solution [184,185].It combines an optical fiber sensor based on the Fabry-Perot (F-P) interference principle with a memory cell composed of a resistive random access memory (RRAM) array based on HfO2 [186].This system can collect,compute,and store the status data,and can be extensively deployed in PV stations as edge computing devices and detection equipment [187].

In addition to the above functions,smart materials facilitate efficient power generation of solar panels.By taking an electrocaloric material as an example,the electric moments align with the orientation of the external electric field,rearranging the inner structure of the material,and decreasing its dipolar entropy [174,175].The electrocaloric effect originates from the adiabatic depolarization of the electrocaloric material,which can be used for cooling systems in PV systems,thereby enhancing the conversion efficiency of electrical energy,particularly in regions exposed to extreme heat [176].Therefore,electrocaloric materials are especially advantageous in high-temperature environments such as deserts or areas with intense solar radiation.

In addition,the compounding of holes and electrons can reduce the efficiency of power generation in organic PV devices.In recent years,ferroelectric materials such as polyvinylidene fluoride-trifluoroethylene (PVDFTrFE) have been used to address this problem [177].The integration of ferroelectric polymer layers into devices can generate ferroelectric dipoles,ensuring large and permanent internal electric fields and accelerating charge transport.Yuan et al.[178] demonstrated that the electric field induced by the ferroelectric layer can amount to 50 V/μm. In comparison to electric field bias,the doping of the PVDF-TrFE layer increases the power conversion efficiency by 50%.Given the stable and efficient energy conversion properties of ferroelectric materials,they are particularly advantageous for PV systems [188].

4.2 Aging phenomena of smart materials and insulation materials

Each polymer in smart materials is composed of chemical bonds with specific dissociation energies,such as the energies of C-C,CH3-H,Si-H,and Si-O bonds,which are 3.50,4.40,3.05,and 3.80 eV,respectively [189].Owing to the influence of various factors such as temperature,UV radiation,and electric field,the external energy originating from these factors may disrupt the chemical bonds and facilitate the formation of new ones [190-192].Therefore,it shall be highlighted that the aging of both the PV panel and sensor materials is accelerated by this process.Although smart composite materials,which serve as the core components of sensors,can significantly enhance the intelligent operation of IoT-based green-smart PV systems,these materials may experience degradation phenomena,such as polarization under extreme environmental conditions,which potentially reduces the intelligence of the system.For instance,in dust storms or thunderstorms,intense electric fields are formed,which promote the aging of smart materials [193,194].Pang et al.[195] investigated the aging phenomenon of polyvinylidene fluoride (PVDF),and demonstrated that the optimal molecular structure was disrupted,and the dielectric insulation performance was compromised due to the increase in polarization.In addition,chemical reactions were likely to occur when the PVDF was exposed to intense electric fields.The performance of the insulation materials in PV backsheets is considered an important factor that enables the secure and stable operation of PV panels.However,these materials are susceptible to degradation when exposed to harsh weather conditions.For example,the optical and chemical properties of ethylene vinyl acetate(EVA) may be affected upon exposure to UV radiation and high temperatures [196].The outer surface may seriously deteriorate when the PET is exposed to UV radiation [197].To prevent the performance of these insulation materials from being compromised by external environmental factors,we present two methods for repairing and preventing the degradation of insulation materials.One method involves applying a sealant to the surface of the PV backsheet,whereas the other method involves reinforcing the internal structure of the PV backsheet.Beaucarne et al.[198]proposed a method based on a single-component flowable silicone sealant,which is significantly less expensive than module replacement or off-site repair.With the sealant,the insulation resistance of the PV modules can be restored,and a protective layer can be provided to prevent degradation of the cracked PV backsheet.Another study proposed a method to prevent the degradation of PET by filling it with a filler and additive [199].When the polymer degraded as a result of erosion due to PD,the fillers,such as BaSO4 and TiO2,remained on the PET surface because their PD resistance was more pronounced than that of the polymer.Thus,the filled PET exhibited better PD resistance than the unfilled PET,enhancing the stability of PET.

5 Conclusions

In this paper,we presented a comprehensive review of the challenges faced by IoT-based green-smart PV systems under extreme climatic conditions.The key findings of this review are summarized as follows.With the increasing installed capacity and voltage of individual PV modules,the problem of power degradation in distributed PV systems due to mismatch and aging under extreme climatic conditions has become increasingly prominent,necessitating more refined sensor deployment strategies to adapt to complex environmental conditions.The use of passive sensors for remote monitoring in extreme environments is becoming more significant and must advance in parallel with the development of new energy harvesting technologies.The variability of extreme climatic conditions may lead to load shedding and power failure,whereas sensor failure or aging can cause data anomalies,highlighting the critical role of IoT in sensor interconnection.Consequently,big data analysis and learning algorithms based on the IoT must be continuously iterated and optimized to enhance system adaptability and reliability.

Based on the aforementioned problems,we thoroughly explored the strategies and technological advancements to address the challenges posed by extreme climatic conditions from various perspectives,including PV and new energy planning and design,aging effects under harsh environmental conditions,optimization of data processing and analysis methods,development of standardized testing procedures,lifetime prediction,and failure mechanisms,as well as environmental protection and recycling.

6 Future prospects

Currently,there is a pressing need to develop an integrated sensor framework,particularly for extreme climatic conditions.In desert regions,for example,dust accumulation on PV surfaces can lead to energy losses of up to 80%.Moreover,in areas such as desert regions,the optimal lifetime of PV panels is limited to 10 years because of the erosive effects of dust accumulation and other factors,simultaneously increasing the costs associated with the frequent replacement of damaged components.

Besides,the aging issue of photovoltaic solar panels used in space and on the Moon under extreme climates in the future is a complex and multifaceted challenge that requires comprehensive consideration and analysis from material selection,design optimization,as well as the integration of intelligent sensing IoT technology.

The use of an integrated sensor framework is critical in such scenarios.With IoT-based sensor systems,real-time monitoring can be conducted effectively.In addition,the angles of the PV panels can be automatically adjusted to achieve optimal performance.Furthermore,the underlying faults can be identified,providing smart energy systems with preventive maintenance.Such systems will benefit future renewable energy systems in terms of efficiency and reliability,making the world greener and more sustainable.

Acknowledgments

This work is supported by the National Key R&D Program of China (Grant No.2023YFE0114600),The National Natural Science Foundation of China (NSFC)-(Grant No.52477029),Joint Laboratory of China-Morocco Green Energy and Advanced Materials,The Youth Innovation Team of Shaanxi Universities,The Xi’an City Science and Technology Project (No.23GXFW0070) and Xi’an International Science and Technology Cooperation Base.

Declaration of Competing Interest

We declare that we have no conflict of interest.