Real Photovoltaic System Research Articles

Deep Reinforcement Learning using Deep-Q-Network for Global Maximum Power Point Tracking: Design and Experiments in Real Photovoltaic Systems

This paper presents a methodology for integrating Deep Reinforcement Learning (DRL) using a Deep-Q-Network (DQN) agent into real-time experiments to achieve the Global Maximum Power Point (GMPP) of Photovoltaic (PV) systems under various environmental conditions. Conventional methods, such as the Perturb and Observe (P&O) algorithm, often become stuck at the Local Maximum Power Point (LMPP) and fail to reach the GMPP under Partial Shading Conditions (PSC). The main contribution of this work is the experimental validation of the DQN agent’s implementation in a synchronous DC-DC Buck converter (step-down converter) un- der both uniform and PSC conditions. Additionally, we establish a testing pipeline for DRL models. The DQN agent’s performance is evaluated alongside the P&O algorithm. Results consistently indicate the DQN agent’s superiority over the P&O algorithm in all simulated scenarios. Although this trend does not entirely replicate in real-world test setups, significant results are observed. Specifically, in PSC sce- narios where the P&O algorithm becomes trapped at an LMPP, the DQN algorithm extracts up to 63.5% more power than the P&O algorithm. An open repository is available, containing PCB schematic designs and layouts, along with the code used for model training and deployment.

A Novel EPSO Algorithm Based on Shifted Sigmoid Function Parameters for Maximizing the Energy Yield from Photovoltaic Arrays: An Experimental Investigation

This paper presents a new method for maximizing the energy production of photovoltaic (PV) arrays, utilizing the Enhanced Particle Swarm Optimization (EPSO) algorithm. In contrast with the classical PSO, the proposed EPSO algorithm employs shifted sigmoid functions in order to adapt the acceleration coefficients and the inertia weight instead of using constant coefficients. The EPSO algorithm proves its effectiveness even in challenging either condition, such as abrupt variations in solar irradiance or instances of partial shading conditions (PSCs). The proposed EPSO-based Maximum Power Point Tracking (MPPT) controller has undergone testing and a comparative analysis alongside well-known metaheuristic algorithms, which include the PSO, Bat Algorithm (BAT), Grasshopper Optimization Algorithm (GOA), and Grey Wolf Optimizer (GWO). The results obtained through simulations consistently demonstrate that the EPSO-based MPPT method surpasses other metaheuristic approaches in terms of energy yield and the speed at which it tracks the Maximum Power Point (MPP) under PSCs. In addition, the proposed EPSO algorithm demonstrates robustness, maintaining consistent tracking of the MPP during sudden solar radiation changes. Experimental validation on a real PV system affirms these findings, demonstrating that the EPSO algorithm surpasses its counterparts by achieving a high MPPT efficiency of approximately 99.19% in a fast-tracking time of 2.4 s while maintaining minimal power oscillations of around 2.04 %.

Empirical Characterization of Non-Intentional Emissions generated by PV Installations

Non-Intentional Emissions (NIEs) generated by solar photovoltaic (PV) systems have significantly increased concerns about disturbances in the 9-150 kHz band of the lowvoltage distribution grid, which can lead to failures or cause thermal stress in the devices connected to the grid. Moreover, these emissions can cause interference in narrowband power line communications (NB-PLCs). This paper presents a characterization of emissions generated by different real PV systems. Specifically, measurement campaigns were carried out in isolated PV systems, which employed different panel configurations, as well as in a grid-connected system. The analysis of the NIEs allows the identification of the frequencies and amplitudes of the emissions, as well as the evaluation of the NIEs variation over time. In addition, the impedances of the PV systems were also measured to correlate impedance variations with emissions.

Anomaly detection in PV systems using constrained low-rank and sparse decomposition

PV (photovoltaic) systems, also known as solar panel systems, play an essential role in the mitigation of greenhouse gas emissions and the promotion of renewable energy. Through the conversion of sunlight into usable energy, electricity is generated without emitting greenhouse gases and producing pollutants. Notwithstanding the evolutionary significance of PV systems, the occurrence of defects and anomalies in PV systems may result in diminished power output, consequently impeding the efficiency of the systems and potentially resulting in hazards in certain circumstances. Therefore, early detection of faults and anomalies in PV systems is imperative to guarantee the reliability, efficiency, and safety of the systems. In this article, we develop a signal decomposition for the purpose of anomaly detection in PV systems. The proposed methodology is grounded on the concept of low-rank and sparse decomposition, with consideration given to the signs of the decomposed low-rank and sparse components, as well as the smooth variations within and between periods in the mean signals. Through the implementation of Monte Carlo simulations, we showcase the efficacy of our proposed methodology in identifying anomalies of varying durations and magnitudes in PV systems. A case study is employed to validate the proposed methodology in detecting anomalies in real PV systems.

Analysis and Performance Comparison of IoT Message Transfer Protocols Applying in Real Photovoltaic System

The adoption of reliable and real-time communication technology is an absolute necessity for the advancement of Internet of Things (IoT) applications. Messaging protocols such as MQTT, AMQP, and HTTP are frequently used for communication with resource-constrained IoT devices. However, choosing a suitable and effective messaging protocol presents a daunting challenge for organizations, as it depends on the specific characteristics and messaging requirements of the IoT system. Therefore, it is crucial to have a comprehensive understanding of three established messaging protocols, such as the Hypertext Transfer Protocol (HTTP), the Message Queuing Telemetry Transport (MQTT), and the Advanced Message Queuing Protocol (AMQP), to appropriately apply them in practical projects. In this paper, information technology solutions are provided for a chain of solar farms to improve harvest productivity, facilitate warning notifications, and enable remote control. Subsequently, a detailed comparative analysis is performed, considering various interconnected criteria, to gain valuable insight into the strengths and limitations of these protocols. The results show that MQTT and AMQP play a role in enhancing overall efficiency and speed within the framework of our suggested photovoltaic system.

Extraction of maximum power from PV system based on horse herd optimization MPPT technique under various weather conditions

This paper introduces a novel approach that extracts the maximum power from photovoltaic (PV) system utilizing the Horse Herd Optimization (HHO) algorithm under different weather conditions, including fast changes in solar radiation (FCSR) and partial shading conditions (PSCs). The HHO algorithm is a technique for optimization that mimics the movement behavior exhibited by horses within a herd. The proposed MPPT controller has been tested and compared with other recognized metaheuristic algorithms, including the Grey Wolf Optimizer (GWO), Particle Swarm Optimization (PSO), Flower Pollination (FP), Deterministic PSO (DPSO), and Cuckoo Search (CS). According to the simulation results, the proposed HHO-based MPPT method is found to outperform other considered metaheuristic methods in terms of the maximum power extraction, fast-tracking, and settling times under various weather conditions. Additionally, the suggested MPPT controller is robust and can continuously track the MPP under the FCSR. The performance of the proposed HHO controller is validated experimentally on a real PV system. It is demonstrated that the proposed HHO algorithm is robust and capable of outperforming the other counterpart metaheuristic algorithms as well as offering the highest MPPT efficiency of about 98.81 % with a rapid tracking time of 2.2 s. Furthermore, it has the lowest power oscillations of about 1.91 %, ensuring stable and consistent power output.

Selection of the photovoltaic system power for the electric vehicle

The article presents a method of a photovoltaic system power selecting for proper electric vehicle. At the beginning, the characteristics of the electric vehicle itself and its traction batteries were made in order to determine the method of its charging. In the presented studies, measurement data from the operation of real photovoltaic systems was recorded and processed. The algorithm used includes examining the energy consumption profile of the owner's residential building. The combined demand for electricity of the electric vehicle and the building made it possible to determine the planned photovoltaic system. The authors presented three possibilities of its location. It can be installed on the roof of the building, on the ground next to the building or on the carport under which an electric vehicle can be parked. Finally, the Metalog family of probability distributions was used to analytically validate the power choice of the photovoltaic system. The authors have developed an algorithm using human and artificial intelligence that helps to properly select the power of the photovoltaic system for the vehicle.

Partial shading detection and hotspot prediction in photovoltaic systems based on numerical differentiation and integration of the P − V curves

Hotspot phenomenon is an expected consequence of long-term partial shading condition (PSC), which results in early degradation and permanent damage of the shaded cells in the photovoltaic (PV) systems. Hence, partial shading detection and estimation of its intensity is a significant factor to predict the possibility of hotspotting process and its occurrence time for PV system protection. In this paper, two partial shading detection indices are introduced based on the extracted power-voltage ( P − V $P - V$ ) curves. The first P − V $P - V$ characteristic is related to the real shaded PV system and the second P − V $P - V$ curve is numerically obtained from the equivalent single-diode circuit model. The proposed PSC detection indices are established according to the difference between the numerical differentiation and integration of the P − V $P - V$ curves from the real PV system and the virtual reference model. These techniques have the advantages to distinguish between temporary and permanent PSC. Moreover, it is possible to estimate the intensity of the shading level and predict the hotspot occurrence time, accordingly. The performance of these methods are confirmed using some simulations and an experimental setup. The results verify the capabilities of the methods for partial shading detection without the requirement of an extra physical cell/module as a reference model.

An enhanced Harris Hawk optimization algorithm for parameter estimation of single, double and triple diode photovoltaic models

Due to the rapid development of photovoltaic (PV) system and spreading of its application, the accuracy of modeling of solar cells, as the main and basic element of PV systems, is gaining relevance. In this paper, an Enhanced Harris Hawk Optimization Algorithm (EHHO) is proposed and applied for estimating the required parameters of different PV models in an effective and accurate way. Harris Hawk Algorithm (HHO) is based on Hawks ways in hunting and catching their preys. The HHO utilizes two phases including exploration and exploitation. The main purpose of proposed enhancement is to improve the second phase of HHO. This enhancement is performed on the exploration phase by fluctuating toward or outward the best optimal solution using sine and cosine functions. Both conventional and proposed algorithms are applied for single, double and triple diode PV models. In order to test the applicability and robustness of proposed algorithm, it is applied for estimating the parameters of different real PV systems and compared with other recent optimization algorithms. The results show that the proposed EHHO is more accurate than conventional HHO and other algorithms.

A Novel Intelligent ANFIS for the Dynamic Model of Photovoltaic Systems

Developing accurate models for photovoltaic (PV) systems has a significant impact on the evaluation of the accuracy and testing of PV systems. Artificial intelligence (AI) is the science of developing machine jobs to be more intelligent, similar to the human brain. Involving AI techniques in modeling has a significant modification in the accuracy of the developed models. In this paper, a novel dynamic PV model based on AI is proposed. The proposed dynamic PV model was designed based on an adaptive neuro-fuzzy inference system (ANFIS). ANFIS is a combination of a neural network and a fuzzy system; thus, it has the advantages of both techniques. The design process is well discussed. Several types of membership functions, different numbers of training, and different numbers of membership functions are tested via MATLAB simulations until the AI requirements of the ANFIS model are satisfied. The obtained model is evaluated by comparing the model accuracy with the classical dynamic models proposed in the literature. The root mean square error (RMSE) of the real PV system output current is compared with the output current of the proposed PV model. The ANFIS model is trained based on input–output data captured from a real PV system under specified irradiance and temperature conditions. The proposed model is compared with classical dynamic PV models such as the integral-order model (IOM) and fractional-order model (FOM), which have been proposed in the literature. The use of ANFIS to model dynamic PV systems achieves an accurate dynamic PV model in comparison with the classical dynamic IOM and FOM.

Spatio-Temporal Graph Neural Networks for Multi-Site PV Power Forecasting

Accurate forecasting of solar power generation with fine temporal and spatial resolution is vital for the operation of the power grid. However, state-of-the-art approaches that combine machine learning with numerical weather predictions (NWP) have coarse resolution. In this paper, we take a graph signal processing perspective and model multi-site photovoltaic (PV) production time series as signals on a graph to capture their spatio-temporal dependencies and achieve higher spatial and temporal resolution forecasts. We present two novel graph neural network models for deterministic multi-site PV forecasting dubbed the graph-convolutional long short term memory (GCLSTM) and the graph-convolutional transformer (GCTrafo) models. These methods rely solely on production data and exploit the intuition that PV systems provide a dense network of virtual weather stations. The proposed methods were evaluated in two data sets for an entire year: 1) production data from 304 real PV systems, and 2) simulated production of 1000 PV systems, both distributed over Switzerland. The proposed models outperform state-of-the-art multi-site forecasting methods for prediction horizons of six hours ahead. Furthermore, the proposed models outperform state-of-the-art single-site methods with NWP as inputs on horizons up to four hours ahead.

IoT-Based PV Array Fault Detection and Classification Using Embedded Supervised Learning Methods

Faults on individual modules within a photovoltaic (PV) array can have a significant detrimental effect on the power efficiency and reliability of the entire PV system. In addition, PV module faults can create risks to personnel safety and fire hazards if they are not detected quickly. As IoT hardware capabilities increase and machine learning frameworks mature, better fault detection performance may be possible using low-cost sensors running machine learning (ML) models that monitor electrical and thermal parameters at an individual module level. In this paper, to evaluate the performance of ML models that are suitable for embedding in low-cost hardware at the module level, eight different PV module faults and their impacts on PV module output are discussed based on a literature review and simulation. The faults are emulated and applied to a real PV system, allowing the collection and labelling of panel-level measurement data. Then, different ML methods are used to classify these faults in comparison to the normal condition. Results confirm that NN obtain 93% classification accuracy for seven selected classes.

Accelerating the simulation of annual bifacial illumination of real photovoltaic systems with ray tracing

SummaryAccurate modeling of bifacial illumination is critical to improve the prediction of the energy yield of bifacial solar systems. Monte Carlo ray tracing is the most powerful tool to accomplish this task. In this work, we accelerate Monte Carlo ray tracing of large solar systems by nearly 90%. Our model achieves root-mean-square error values of 7.9% and 37.2% for the front and rear irradiance compared against single-axis tracking field reference data, respectively. The rear irradiance modeling error decreases to 18.9% if suspected snow periods are excluded. Crucially, our full system simulations show that surrounding ground surfaces affect the rear irradiance deep into the system. Therefore, unit system simulations cannot necessarily ignore the influence of the perimeter of large installations to accurately estimate annual yield. Large-scale simulations involving high-performance supercomputing were necessary to investigate these effects accurately, calibrate our simplified models, and validate our results against experimental measurements.

Availability estimation in photovoltaic generation systems using Timed Petri Net simulation models

In solar photovoltaic (PV) power generation systems, availability impacts directly on annual energy production capacity. In order to reveal availability levels, the system is usually constructed and monitored. Besides to be expensive and time consuming, this approach fails to provide in-advance estimations that could benefit planning, dimensioning, and predictive maintenance. This paper proposes a Petri Net-based model that can anticipate, within a short period of time and acceptable level of accuracy, availability levels for PV systems with different sizes, architectures, generation profiles, and stages of construction. By receiving a set of offline input parameters, the model can be exploited in advance, before actually constructing the real infrastructure to be measured. Simulations show that the system behavior under variability can be anticipated with reasonable accuracy, which tends to be valuable for predictive engineering. A real PV system model illustrates our approach.

Storage battery operation in autonomous photovoltaic systems in Siberia and the Russian Far East. Practical operating experience

This paper presents real-life experience in operating storage batteries in autonomous photovoltaic systems located in Siberia and the Russian Far East. A description is given of the photovoltaic systems’ installed capacity and the technical specifications of the storage batteries in use. Wide discrepancy in the adopted battery technologies leads to the conclusion that, at the design stage, prospective technical solutions are developed based on subjective evaluations rather than on composition modeling​ and optimization outcomes. The described issues that occurred in real photovoltaic systems in Siberia and the Russian Far East confirm this.The paper proposes methodological and technical measures designed to improve the storage battery operating environment in the harsh conditions of Siberia and the Russian Far East. One of such measures is to solve the equipment optimization problem with respect to storage battery categorization. Furthermore, it is necessary to take better account of the battery room’s in-house energy consumption, given the severe Siberian conditions. Part-time operation of storage batteries in photovoltaic systems throughout the year also needs to be considered. Practice shows that this measure considerably increases storage battery lifetime. This paper summarizes the authors’ examination of numerous photovoltaic systems with further analysis of the storage batteries’ operating modes and other technical factors.

Photovoltaic chain operation analysis in condition of partial shading for systems with and without bypass diodes

The paper focuses on the operation analysis of the photovoltaic chain in condition of partial shading. For the needs of this analysis, the model of this PV system was implemented in the MATLAB/Simulink. This model corresponded to the structure of the real photovoltaic system on which experiments were carried out. The model consists of seven PV modules connected in series, where each of them can receive a different irradiance. Two cases were analyzed, the first when each of the seven modules had a bypass diode and the second when there were no bypass diodes in the system. The model was validated by identifying parameters based on two experiments carried out under different conditions, so that the model corresponds to the actual installation on which the experiments were conducted. The operation of the model was analyzed for various configurations of partial shading and the presence or absence of bypass diodes.

Experimental Study on the Effect of Dust Deposition on a Car Park Photovoltaic System with Different Cleaning Cycles

For a decade, investments in solar photovoltaic (PV) systems have been increasing exponentially in the Middle East. Broadly speaking, these investments have been facing tremendous challenges due to the harsh weather in this particular part of the world. Dust accumulation is one the challenges that negatively affects the performance of solar PV systems. The overall goal of this paper is to thoroughly investigate the effect of dust accumulation on the energy yield of car park PV systems. With this aim in mind, the paper presents scientific values for further research and opens the horizon for attracting further investments in solar PV systems. This study is based on a real PV system in the Sultanate of Oman and considers different cleaning cycles for 16 months (from 29 July 2018 to 10 November 2019). Furthermore, four different PV groups were assessed, and the system was monitored under different cleaning frequencies. In general, it was found that dust accumulation has a significant impact; under 29-day, 32-day, 72-day, and 98-day cleaning cycles, the average percentages of energy loss due to soiling were 9.5%, 18.2%, 31.13%, and 45.6%, respectively. In addition, the dust effect has a seasonal variation. The study revealed that dust accumulation has a more negative impact during summer than during winter. During summer, the energy losses due to soiling were 8.7% higher than those during winter. The difference was attributed to different environmental conditions, with high humidity and low wind speed being the main factors that worsen the impact of dust during summer. Based on the findings of this research, a monthly cleaning program is highly recommended in the city of Muscat.

Characterization of DC series arc faults in PV systems based on current low frequency spectral analysis

This work presents an experimental study focused on the characterization of series arc faults in direct current (DC) photovoltaic (PV) systems. The aim of the study is to identify some relevant characteristics of arcing current, which can be obtained by means of low frequency spectral analysis of current signal. On field tests have been carried out on a real PV system, in accordance with some tests requirements of UL 1699B Standard for protection devices against PV DC arc faults. Arcing and non-arcing current signals are acquired and compared and the behavior of a set of indicators proposed by authors is analyzed. Different measurement equipment have been used, in order to study the impact of both measurement transducers and data acquisition systems on proposed indicators effectiveness. Presented results show that the considered indicators are suitable for detecting the arc presence even with commercial devices normally used for smart metering applications.

Regression and Generalized Additive Model to Enhance the Performance of Photovoltaic Power Ensemble Predictors

Photovoltaic (PV) power prediction has a constantly evolving solutions landscape with a myriad of data-driven techniques. Each technique leverages a self-adaptive algorithm that must retrain in intervals, be it each day, week, or season, to avoid the model generalizing poorly because of overfitting, underfitting, or concept drift. This paper aims to improve the generalization capability of PV power predictors such as autoencoders used widely in the industry by introducing feature-enhanced ensemble learning (FEEL) after the feature selection step. This framework uses a combination of nonparametric regression and generalized additive models, and an ensemble of weak regularized multilayer perceptron models. Once trained, the framework can reliably generalize on test data across long time periods without any significant degradation in performance. The proposed framework was validated against the baseline autoencoder-based feature enhancement model on a real PV system from a smart neighborhood in Alabama for September 2019. The FEEL framework performed three times better than the baseline, but when applied to the baseline, its performance improved by two times on average. Furthermore, the framework generalized consistently better than five other feature enhancement strategies. Despite fluctuations in weather, the FEEL framework’s R-square score had a range of 8.1%, whereas that of the baseline was 48.3%. The mutual information and Minkowski distance scores attempted to quantify concept and model drift, respectively. These scores show that the FEEL framework generalized the ensemble learning models at least two times better than the baseline across the different test days. These results form the first step toward decentralized intelligence for smart grid applications that could free up resources for other expensive analytics in the field.

Efficiency Improvement Scheme for PV Emulator Based on a Physical Equivalent PV-Cell Model

Recently, a photovoltaic (PV) emulator is proposed which is based on a combination of a constant current source and a one-diode photovoltaic model. Its superior dynamic performance is compatible with that of a real PV system. Although it is power efficient at the maximum power point (MPP), it suffers from high power loss around and at the open-circuit voltage (OCV) operation condition. The PV emulator can be used for PV system analysis and testing, such as maximum power point tracking (MPPT). This paper presents a new switching circuit which is placed in parallel with the diode string to minimize the power loss. The switching circuit consists of a two-switch non-inverting buck-boost DC/DC converter. When the operating point of the PV emulator moves from the current source region to the voltage source region, the converter, which is more efficient, switches in to replace the diode string seamlessly to maintain the circuit operation of the emulator. Experimental results show that in the worst case scenario, i.e. OCV condition, the efficiency and temperature of the proposed solution reach 81.47% and 30.1 °C respectively, as compared with 2.8% and 94.2 °C respectively for the diode string only case. In terms of dynamic response, the proposed PV emulator lags the real PV panel by only 3.5 ms as compared with 120 ms by a commercial emulator under the 30% to 60% insolation change test.