Maximum power point tracking (MPPT) technology plays a key role in improving the energy conversion efficiency of photovoltaic (PV) systems, especially when multiple local maximum power points (LMPPs) occur under partial shading conditions (PSC). It is necessary to modify the operating point efficiently and accurately with the help of MPPT technology to maximize the collected power. Even though a lot of research has been carried out and impressive progress achieved for MPPT technology, it still faces some challenges and dilemmas. Firstly, the mathematical model established for PV cells is not precise enough. Second, the existing algorithms are often optimized for specific conditions and lack comprehensive adaptability to the actual operating environment. Besides, a single algorithm may not be able to give full play to its advantages. In the end, the selection criteria for choosing the suitable MPPT algorithm/converter combination to achieve better performance in a given scenario is very limited. Therefore, this paper systematically discusses the current research status and challenges faced by PV MPPT technology around the three aspects of MPPT models, algorithms, and hardware implementation. Through in-depth thinking and discussion, it also puts forward positive perspectives on future development, and five forward-looking solutions to improve the performance of PV systems MPPT are suggested.

Photovoltaic (PV) power is one of the most representative renewable energy resources, which is not only environmentally friendly but also sustainable and expandable [

With the continuous progress of technology, it is worth noting that MPPT technology for PV systems still faces some challenges and dilemmas. First of all, under dynamic operating conditions, the nonlinear characteristics of PV systems become more complex. Over-idealized models such as single diode models (SDM) may not accurately capture the dynamic behavior of the system in real operation, pending optimization or transition to more accurate models. Second, despite the rich diversity of existing MPPT algorithms, most studies have only conducted simulation experiments. They also mainly optimize for specific conditions, lacking universality and comprehensive adaptability to the practical operating environment. Additionally, a single algorithm may not be sufficiently advantageous in the face of increasing demands and system complexity. How to efficiently combine algorithms under given conditions is also an issue worth discussing [

The accurate modeling is essential for the implementation of MPPT techniques in PV systems, several PV cell models have been proposed in previous literature [

SDM simplifies the PV cell as a diode with forward voltage and output current. The light-induced electron-hole pair formation, voltage generation, and current flow are described. Since fewer unknown parameters need to be estimated in SDM, it is used to analyze the battery performance under static conditions in most. DDM considers the reverse saturation current. Whereas, TMD better considers effects like non-uniformity and temperature. While DDM has more parameters and higher computational complexity, it is closer to the actual situation and can more accurately describe the

References [^{−4}, and 9.0 * 10^{−4}, respectively. The SDM-based MPPT algorithm also had the lowest RMSE of all the compared algorithms. Reference [^{2} and a temperature of 25°C are 2 ms and 99.88%, respectively. Both were the best among all the compared algorithms, while the SDM-based one required 135 ms and the efficiency was only 98.70%. The comparison in

Model | Equivalent circuit | Output current & output voltage | Remarks | |
---|---|---|---|---|

Advantages | Disadvantages | |||

Single-diode |
Simple and accessible | Insufficient precision | ||

High calculating efficiency | Constrained by low irradiance |
|||

Low cost | ||||

Double-diode |
More accurate | More complex | ||

Better adapted to dynamic changing conditions | Performance depends on Model parameters | |||

Wider range of application scenarios | ||||

Moderate cost | ||||

Triple-diode |
More comprehensive modeling | Computation complexity | ||

More precision | More difficult to adjust parameters | |||

Improved response to dynamic operating conditions | Unnecessary complexity due to excessive detail | |||

High cost |

At present, the research and development of MPPT algorithms for PV systems mainly focus on several directions, including traditional algorithms, optimization algorithms, intelligent algorithms, and hybrid algorithms [

The conventional algorithms are more comprehensive in terms of basic research as well as simple and easy to implement. They are more efficient in the case of uniform irradiance. The most common conventional algorithms include perturb and observe (P&O) [

P&O and HC both determine the direction in which the optimal output power changes by introducing a perturbation. The only difference between them is the perturbation parameter. To track the MPP, the P&O senses and perturbs the voltage or current, while the HC disturbs the duty cycle. Moreover, the INC is used to determine the optimum operating point by detecting the ratio of conductance derivative to instantaneous conductance over time. CV uses a reference voltage or a value fixed under specific conditions to control the MPP voltage of the PV system. Similarly to CV, CC makes the system operate at a constant current state. The comparison of different conventional algorithms is shown in

Algorithm | Tracking speed | Tracking accuracy | Efficiency | Complexity | Economy | Sensed parameters | Features | |
---|---|---|---|---|---|---|---|---|

Advantages | Disadvantages | |||||||

P&O | Medium | High | Medium | Medium | Moderate | Simple control structure | Higher steady-state power losses | |

Few measurement parameters | Oscillations around MPP | |||||||

High tracking capabilities | Local optimum under PSC | |||||||

INC | Medium | High | Medium | Medium | Moderate | Less power losses | Easily fall into local optimization | |

Low oscillations frequency around MPP | Variable step size requires complex and costly |
|||||||

CV | Slow | Low | Low | Low | Cheap | Simple and easy to implement | Poor tracking capabilities | |

High stability | High power losses | |||||||

Adapted to small temperature changes | Local optimum under PSC | |||||||

Dependent on PV module characteristics | ||||||||

CC | Slow | Low | Low | Low | Cheap | Simple and easy to implement | Poor tracking capabilities | |

High stability | High power losses | |||||||

Adapted to small temperature changes | Local optimum under PSC | |||||||

Dependent on PV module characteristics | ||||||||

HC | Medium | Medium | Medium | Low | Moderate | Simplicity and easy Implementation | Low accuracy | |

Independent of PV module characteristics | Difficulty in trading off performance between steady state and dynamic response error |

Optimization algorithms inspired by natural processes or behavioral patterns of biologists find the best solution to a problem from all possible alternatives. They are less likely to fall into a local optimum under dynamic operating conditions and can be more flexible in solving nonlinear problems [_{c}). Individuals associated with larger values of fitness indicate that it has a better or higher quality solution. The fitness values for each control cycle will be determined by collecting the actual voltage and current.

Algorithm | Tracking speed | Tracking accuracy | Efficiency | Complexity | Economy | Sensed parameters |
---|---|---|---|---|---|---|

PSO | High | Medium | Very high | Medium | Moderate | |

GA | Medium | High | High | High | High | |

GWO | Medium | High | Very high | Medium | Moderate | |

ACO | High | Medium | Very high | Low | Moderate | |

FA | Medium | High | High | Medium | Moderate |

Intelligent algorithms have the potential to solve many of the problems associated with conventional MPPT methods without complex arithmetic or precise parameters, such as increasing tracking speed, reducing computation time, and minimizing power fluctuations near the GMPP. Widely used intelligent algorithms include fuzzy logic control (FLC) [

Algorithm | Tracking speed | Tracking accuracy | Efficiency | Complexity | Economy | Sensed parameters |
---|---|---|---|---|---|---|

FLC | High | High | High | Low | Moderate | |

SMC | Very high | Medium | High | High | Expensive | |

ANN | Medium | High | High | Medium | Expensive | |

Gauss-newton | High | Medium | High | High | Moderate |

In the face of growing demands and system complexity, researchers have gradually realized that a single algorithm may not be able to give full play to its advantages. Therefore, hybrid algorithms have gradually become the mainstream trend, achieving more comprehensive and superior MPPT performance through the synergistic cooperation of different algorithms. There are three combinations of hybrid algorithms, which are traditional and traditional algorithms, traditional and intelligent or optimization algorithms, and intelligent and optimization algorithms. Reference [

Algorithm | Tracking speed | Tracking accuracy | Efficiency | Complexity | Economy | Sensed parameters |
---|---|---|---|---|---|---|

INC-IBSC | High | Medium | High | Medium | Moderate | |

ARO-P&O | High | High | High | Medium | Moderate | |

GS-PS-ANN-P&O | Very high | Very high | Very high | Very high | Very expensive |

DC-DC converter is one of the important hardware conditions for realizing MPPT in PV systems. MPPT algorithms generally work together with DC-DC converter to optimize the performance of PV systems and maximize the energy conversion efficiency under different operating conditions [

Reference [

MPPT technology for PV systems has made noteworthy progress in existing research, but still faces a series of challenges and opportunities:

High-precision modeling for PV systems will bring about high complexity and high-cost issues. In the future, a reasonable balance between model accuracy, cost, and system complexity should be considered. In addition, it was discovered during the discussion that the number of studies applying TMD to the PV systems MPPT was currently limited. This can be extended and explored more.

The conventional MPPT methods are simple in principle and hardware implementation, but they have low accuracy, slow speed, and tend to oscillate around the MPP. They are suitable for MPPT under uniform conditions. Intelligent algorithms do not require precise mathematical models and have high tracking efficiency. However, the realization cost is high and not widely used in PV MPPT. Optimization algorithms are applicable to complex nonlinear systems. Especially under PSC, it is the best choice to realize MPPT. However, they are highly stochastic and more dependent on the parameter settings. It will take a lot of time and effort to continuously tune the parameters.

Hybrid algorithms integrate the advantages of different algorithms, and their validation has been verified. Such algorithms suffer from the disadvantages of high complexity and long development cycles. With the help of deep learning, real-time data mining, multi-converter fusion, and other techniques, further exploration and research can be carried out on how to choose the right combination of algorithms for specific operating conditions.

Although many algorithms have been proposed to realize MPPT in PV systems, they only stay in the laboratory environment and simulation testing stage. When applied to large-scale PV systems, some algorithms might lead to fluctuations in the operating point, reducing system stability. Hardware-in-loop (HIL) is often used to evaluate the hardware implementation ability of a system. It puts part of the system hardware into the software simulation loop, which can compensate for the absolute idealization of purely digital simulation and improve the confidence of the whole model. By utilizing HIL, researchers are able to further validate the MPPT performance of various algorithms in real applications.

Despite many comparative studies on DC-DC converters and MPPT algorithms in the field of PV systems, the selection criterion that defines the selection of the suitable combination of MPPT algorithms/converters for PV systems to achieve better performance is very limited. That could be achieved by using a multi-objective optimization approach, which integrates evaluation metrics, e.g., the efficiency, stability, and response time of the PV systems. The fuzzy mathematics can be used to calculate the satisfaction of the Pareto optimal frontier solution and select the solution with the largest satisfaction as the compromise optimal solution. The weights are taken to obtain the final score of each combination. A potential advantage of multi-objective optimization is the ability to reveal interrelationships between different optimization objectives, such as trade-offs or conflicts that may exist between certain objectives.

This paper provides a systematic discussion of the current research status and challenges faced by PV MPPT technologies around the three aspects of MPPT models, algorithms, and hardware implementation. It also puts forward positive perspectives on future development. Specifically, the major conclusions claimed in this paper are as follows:

Improving the accuracy of PV systems models requires a reasonable balance between model complexity and economic cost, rather than blindly increasing accuracy. By comparing SDM, DDM, and TMD for PV systems, it was found that DDM had faster tracking speed and efficiency than SDM, which could be used as the best choice for more accurate modeling of PV systems MPPT.

Four types of algorithms are studied and discussed. Conventional MPPT algorithms are suitable for uniform conditions. Intelligent algorithms do not require accurate mathematical modeling and tracking efficiency is high. Optimization algorithms under PSC are the best choice to implement MPPT. Two latest optimization algorithms that have not yet been applied to PV MPPT are provided. Furthermore, hybrid algorithms are more effective in PSC and rapidly changing environmental conditions. Combining multiple algorithms builds on strengths and avoids weaknesses.

Various MPPT algorithms for PV systems have been developed, but most of them are only simulation experiments. Nevertheless, it is also necessary to compensate for the absolute idealization of purely digital simulations with the help of HIL experiments. Evaluate the hardware implementation ability of the system. Especially under dynamic operation and large-scale application conditions.

Hybrid algorithms have great potential to select the right combination of algorithms for specific operating conditions with the help of techniques such as deep learning, real-time data mining, and multi-converter fusion.

Non-isolated DC-DC converters still dominate. Buck-boost converters are characterized by high performance and low power losses, making them the best choice for low power loads. Besides, the FLC/Cuk combination always shows excellent efficiency performance, no matter how the irradiance, temperature, or load varies. This combination has great potential. By integrating the evaluation metrics of efficiency, stability, and response time of the PV systems using a multi-objective optimization approach. It is expected to help find the MPPT algorithms/DC-DC converters combination with optimal performance.

PV system output current (A)

PV system output voltage (V)

PV system output power (W)

Instantaneous irradiance (W/m^{2})

Temperature (°C)

Hardware-in-loop

The authors would like to thank and acknowledge the Intelligent Electric Power Grid Key Laboratory of Sichuan Province.

The authors received funding from the Open Fund Project of Intelligent Electric Power Grid Key Laboratory of Sichuan Province under Grant (2023-IEPGKLSP-KFYB03) and Yunnan Provincial Basic Research Project (202301AT070443).

The authors confirm contribution to the paper as follows: Conceptualization, methodology, resources, validation: Bo Yang; investigation, data collection, writing-original draft and editing, analysis and interpretation of results: Rui Xie; visualization and contributed to the discussion of the topic: Zhengxun Guo. All authors reviewed the results and approved the final version of the manuscript.

The authors confirm that the data used in this study are available on request.

The authors declare that they have no conflicts of interest to report regarding the present study.