In the operation and maintenance management of mountainous centralized photovoltaic power stations, UAV drones technology is gradually replacing traditional manual inspection methods. Previously, power station maintenance and inspection relied heavily on human labor. Although monitoring devices could provide some information, there were still many limitations. UAV drones, with their high-altitude operational capabilities, are particularly suitable for vast mountainous power stations. They not only significantly reduce human and material resource consumption but also greatly improve the timeliness and stability of operation and maintenance. This article delves into the path planning of UAV drones for photovoltaic inspections, taking a specific photovoltaic power station as an example. The research focuses on the rationality of area division, the efficiency of UAV drones path planning, and the comprehensive application of vehicle-drone collaborative path planning. By optimizing these key aspects, the aim is to enhance the efficiency and accuracy of UAV drones inspections, ensuring the safe and stable operation of photovoltaic power stations. This study not only provides theoretical support for the application of UAV drones in photovoltaic inspections but also contributes practical wisdom to the sustainable development of the photovoltaic industry.
The adoption of UAV drones for inspection represents a paradigm shift in the renewable energy sector. The ability of UAV drones to cover large areas quickly and capture high-resolution imagery makes them indispensable for modern photovoltaic farms. However, the effectiveness of UAV drones inspections heavily depends on optimized path planning. Without careful planning, UAV drones may waste energy, miss critical components, or even collide with obstacles. Therefore, developing robust path planning algorithms is crucial. This paper presents a comprehensive framework for UAV drones path planning, incorporating area division, route optimization, and vehicle-drone collaboration. The methodologies discussed here are designed to handle the unique challenges posed by mountainous terrain, such as irregular layouts, elevation changes, and limited accessibility.
The case photovoltaic power station covers an area of 1116 acres, with an installed capacity of 45 MW. It achieved full-capacity grid connection on September 28, 2024, with 15 box transformers successfully put into operation. The project innovatively adopts a “photovoltaic +” model, leveraging local abundant solar energy resources and sufficient land resources. This not only realizes efficient utilization of photovoltaic power generation but also promotes diversified land use and increases agricultural income. After operation, the power station’s annual power generation is approximately 78.69 million kWh, with an estimated annual output value of 22.03 million yuan. Due to the presence of large installations and numerous uncertain factors within the power station, manual inspections are not only time-consuming and labor-intensive but may also pose threats to the safety of inspection personnel. To address this, UAV drones play a vital role in the inspection of mountainous centralized photovoltaic power stations. With their aerial perspective and precise positioning, UAV drones can promptly identify equipment failures, foreign object遮挡, and other issues, ensuring the safe and stable operation of the power station. This inspection method not only improves work efficiency but also reduces labor costs, providing strong support for the long-term maintenance of the power station and guaranteeing its continuous and efficient production of green energy.
The path planning problem for UAV drones inspection can be systematically decomposed into four interrelated key technical modules, as shown in the conceptual framework. These four technical dimensions collectively construct the theoretical framework and technical system for UAV drones inspection path planning, providing important methodological support for achieving intelligent and efficient inspection of photovoltaic power stations. The modules include area division, path planning for individual partitions, optimization of partition visitation order, and vehicle-drone collaboration. Each module addresses specific challenges and contributes to the overall efficiency and safety of UAV drones operations. In the following sections, I will detail each module, presenting mathematical models, algorithms, and practical implementations.
Area Division for UAV Drones Inspection
The case photovoltaic power station is scientifically divided into three areas to facilitate more effective management and inspection. To more efficiently organize and analyze these photovoltaic strings, this paper employs the minimum bounding rectangle method to define the rectangular range of each string. This ensures accurate capture of each string’s boundaries, thereby facilitating subsequent area division and inspection planning. To precisely describe the position and orientation of each photovoltaic string, a feature vector is introduced:
$$ p_i = (x_i, y_i, \theta_i) $$
where \( x_i \) and \( y_i \) are the horizontal and vertical coordinates of photovoltaic string \( i \) on a two-dimensional plane, respectively; \( \theta_i \) is the orientation angle of the string, a parameter crucial for understanding how the photovoltaic panels receive sunlight. In the division process, it is also necessary to calculate the distance \( d(p_i, p_j) \) between two photovoltaic strings. This distance is comprehensively calculated based on their coordinates on the two-dimensional plane and orientation angles. It not only reflects the physical distance but also considers the impact of string orientation on solar energy utilization. By precisely calculating this distance, inspection paths can be arranged more reasonably, ensuring each string receives sufficient attention and timely maintenance. The calculation formula is as follows:
$$ d(p_i, p_j) = \sqrt{(x_i – x_j)^2 + (y_i – y_j)^2} + \alpha |\theta_i – \theta_j| $$
where \( \alpha \) is a weighting factor that balances the influence of Euclidean distance and angular difference. The orientation angle \( \theta_i \) is measured in radians, and the term \( \alpha |\theta_i – \theta_j| \) penalizes strings with vastly different orientations, as they may require different inspection approaches due to varying sunlight exposure.
Next, the density-based spatial clustering of applications with noise (DBSCAN) algorithm is used for corner area division, with key parameters being the neighborhood radius \( \epsilon \) and the minimum number of points in the neighborhood \( \text{MinPts} \). For photovoltaic strings within corner areas, an improved agglomerative hierarchical clustering algorithm is employed for further subdivision. Let \( S \) be the set of all photovoltaic strings, then each string \( p_i \in S \) can be clustered into belonging to a certain partition \( C_k \). The following table presents the partitioned photovoltaic string situation based on the above method.
| Partition ID | Corner Area ID | Number of Strings in Area | Average Position Coordinates | Average Orientation Angle | Neighborhood Radius \( \epsilon \) | Minimum Points \( \text{MinPts} \) |
|---|---|---|---|---|---|---|
| 1 | 1 | 12 | (100, 200) | 30° | 50 | 5 |
| 2 | 2 | 15 | (150, 250) | 45° | 40 | 4 |
| 3 | 3 | 10 | (200, 300) | 60° | 30 | 3 |
The content of Table 1 describes the situation of several partitions formed after area division. Each partition contains a certain number of photovoltaic strings and has average position coordinates and average orientation angles. The neighborhood radius and minimum number of points in the neighborhood reflect the DBSCAN parameters used in the clustering process. This information helps UAV drones plan efficient flight paths according to partitions during inspection.
The DBSCAN algorithm is particularly effective for clustering spatial data with noise. It defines clusters as dense regions of points separated by low-density regions. For photovoltaic strings, this means grouping strings that are geographically close and have similar orientations. The parameters \( \epsilon \) and \( \text{MinPts} \) control the density threshold. A larger \( \epsilon \) or smaller \( \text{MinPts} \) results in fewer but larger clusters, while a smaller \( \epsilon \) or larger \( \text{MinPts} \) leads to more but smaller clusters. In mountainous terrain, where strings may be scattered due to topographical constraints, tuning these parameters is essential for meaningful partitions. The improved agglomerative hierarchical clustering further refines these partitions by merging or splitting clusters based on a linkage criterion, such as average linkage or Ward’s method.
To validate the partitioning results, we can compute the silhouette score for each partition. The silhouette score measures how similar an object is to its own cluster compared to other clusters. For a partition \( C_k \), the silhouette score \( s(i) \) for string \( i \) is defined as:
$$ s(i) = \frac{b(i) – a(i)}{\max\{a(i), b(i)\}} $$
where \( a(i) \) is the average distance from string \( i \) to other strings in the same partition, and \( b(i) \) is the minimum average distance from string \( i \) to strings in any other partition. The overall silhouette score for the partitioning is the average of \( s(i) \) over all strings. A score close to 1 indicates well-separated clusters, while a score near 0 suggests overlapping clusters. For the partitions in Table 1, assuming typical distances, the silhouette score might range from 0.6 to 0.8, indicating good separation.
Path Planning for UAV Drones Inspection
After completing the area division for the case photovoltaic power station, the path planning for UAV drones inspection within a single partition becomes particularly important. Considering the specific layout of photovoltaic strings, this paper assumes all strings are rectangular with consistent aspect ratios. On this basis, path planning needs to comprehensively consider string positions, dimensions, and the field of view of the camera mounted on the UAV drones. Therefore, when planning the flight path of UAV drones, precise calculations must be performed based on the distribution and dimensions of photovoltaic strings, combined with the camera’s field of view width and height, as well as the flight altitude. By optimizing the path, it is ensured that UAV drones can efficiently and accurately complete the inspection task for the entire partition, thereby improving inspection efficiency and accuracy, and providing strong support for the safe and stable operation of the power station.
In specific planning, the center coordinates \( (x_i, y_i) \) of each photovoltaic string are known, as well as its width \( w_i \) and height \( h_i \). Simultaneously, key parameters include the camera’s field of view width \( W \), height \( H \), and flight altitude \( h \). To ensure the images captured by UAV drones are complete and comprehensive, path design must satisfy certain conditions: during flight, the camera’s field of view must fully cover each photovoltaic string, avoiding omissions or repeated shots.
$$ W \geq w_i + 2 \cdot \delta_x $$
$$ H \geq h_i + 2 \cdot \delta_y $$
where \( \delta_x \) and \( \delta_y \) are safety margins in the horizontal and vertical directions, respectively. These margins account for potential positioning errors of UAV drones and ensure that the entire string is within the camera’s field of view even if the UAV drones deviates slightly from its planned path. The safety margins can be expressed as functions of flight altitude \( h \) and camera parameters:
$$ \delta_x = h \cdot \tan\left(\frac{\phi_x}{2}\right) $$
$$ \delta_y = h \cdot \tan\left(\frac{\phi_y}{2}\right) $$
where \( \phi_x \) and \( \phi_y \) are the horizontal and vertical angular fields of view of the camera, respectively. To plan the path, we also need to consider the distance between adjacent strings. Let the horizontal distance between adjacent strings be \( d_x \), and the vertical distance be \( d_y \). Then the movement path of UAV drones can be represented as:
$$ L = M \times (N – 1) \times d_x + (M – 1) \times d_y $$
where \( L \) is the movement path of UAV drones; \( M \) is the number of rows of photovoltaic strings; \( N \) is the number of columns of photovoltaic strings; \( d_x \) is the horizontal distance between adjacent strings; \( d_y \) is the vertical distance between adjacent strings. To adapt to photovoltaic strings of different sizes, we adjust the sizes of \( d_x \) and \( d_y \) to match the actual dimensions of the strings. This can be achieved through the following formulas:
$$ d_x = w_i + m $$
$$ d_y = h_i + n $$
where \( m \) and \( n \) are safety distances determined based on the camera’s field of view and flight altitude. To find the optimal partition inspection sequence, a genetic algorithm can be used. Let \( P \) be the set of partitions, and \( O \) be a visitation order of partitions. Then the optimization objective function can be represented as:
$$ \min \sum_{i=1}^{|P|-1} \text{dist}(p_{O(i)}, p_{O(i+1)}) $$
where \( \text{dist}(p_{O(i)}, p_{O(i+1)}) \) is the flight distance from partition \( p_{O(i)} \) to \( p_{O(i+1)} \). The genetic algorithm evolves a population of candidate sequences over generations, using selection, crossover, and mutation operators to minimize the total travel distance. The fitness of a sequence is inversely proportional to the total distance. The following table presents the parameters determined based on the adaptive path planning method.
| Partition ID | String ID | String Center Coordinates (x, y) | String Dimensions (w, h) | Camera Field of View (W, H) | Horizontal Movement Distance \( d_x \) | Vertical Movement Distance \( d_y \) |
|---|---|---|---|---|---|---|
| 1 | 1 | (100, 100) | (2, 1) | (3, 2) | 2.5 | 1.5 |
| 1 | 2 | (125, 115) | (2, 1) | (3, 2) | 2.5 | 1.5 |
| 2 | 1 | (200, 200) | (2, 1) | (3, 2) | 2.5 | 1.5 |
| 2 | 2 | (225, 215) | (2, 1) | (3, 2) | 2.5 | 1.5 |
| 3 | 1 | (300, 300) | (2.5, 1.5) | (4, 3) | 3.0 | 2.0 |
Table 2 showcases the center coordinates, dimensions, camera field of view sizes, and movement distances per step for photovoltaic strings determined based on the adaptive path planning method. Through such planning, it is ensured that every photovoltaic string can be completely covered and photographed clearly, while guaranteeing the shortest path for UAV drones movement between different strings. This helps improve inspection efficiency and reduce energy consumption.
In addition to the genetic algorithm, other optimization techniques can be applied for path planning. For instance, the traveling salesman problem (TSP) framework can be adapted for partition visitation order. Since TSP is NP-hard, heuristic methods like simulated annealing or ant colony optimization might be employed. Moreover, within a partition, the path for UAV drones can be modeled as a coverage path planning (CPP) problem. The goal of CPP is to ensure that every point in the area of interest is covered by the camera’s field of view. For rectangular strings arranged in grids, a simple boustrophedon (lawnmower) pattern is often optimal. However, in mountainous terrain, strings may not align perfectly, requiring more complex patterns. Algorithms like spanning tree coverage or cellular decomposition can be used to generate efficient coverage paths.
The flight path \( L \) can be further optimized by considering the energy consumption of UAV drones. The energy required for flight is proportional to the distance traveled and the weight of the UAV drones. Let \( E_{\text{total}} \) be the total energy available, and \( e_{\text{unit}} \) be the energy consumed per unit distance. Then the maximum allowable path length is \( L_{\text{max}} = E_{\text{total}} / e_{\text{unit}} \). The path planning must satisfy \( L \leq L_{\text{max}} \). If the path exceeds this limit, the inspection area must be subdivided further, or battery swapping schedules must be integrated.
Vehicle-Drone Collaboration for UAV Drones Inspection
Considering the complex terrain of the case photovoltaic power station located in mountainous areas, this paper proposes a UAV drones and inspection vehicle collaboration scheme. UAV drones are responsible for high-altitude operations, utilizing their flexibility and wide视角 to comprehensively scan the 1116-acre photovoltaic area, quickly identifying equipment abnormalities. Simultaneously, ground inspection vehicles equipped with professional devices conduct detailed checks and confirmations on suspected problems discovered by UAV drones. The two achieve information sharing and collaborative operations through real-time data transmission. This scheme not only improves inspection efficiency and accuracy but also enables timely response to equipment failures, ensuring the safe and stable operation of the power station, maximizing the utilization of solar energy resources, and providing solid assurance for the long-term efficient operation of the power station.
To integrate the aforementioned issues regarding UAV drones endurance management and inspection efficiency under the vehicle-drone collaboration mode, this paper establishes a comprehensive mathematical model to evaluate inspection efficiency and UAV drones safety under different collaboration methods. Define: \( S \) as the partition size; \( v \) as the speed of UAV drones; \( T_{\text{fly}} \) as the single flight time of UAV drones; \( T_{\text{chg}} \) as the single battery更换 time of UAV drones; \( P_d \) as the damage probability of UAV drones in a single task; \( E \) as the inspection efficiency per unit time; \( R \) as the safety indicator for a single task of UAV drones. Then inspection efficiency \( E \) can be expressed as:
$$ E = \frac{S}{T_{\text{eff}}} $$
where \( S \) is the partition size, i.e., the area of the entire inspection region; \( v \) is the speed of UAV drones; \( T_{\text{eff}} \) is the effective inspection time, the total time spent by UAV drones from starting the inspection of the first string to completing the inspection of the last string, excluding non-inspection times such as takeoff, landing, and往返 to the site; \( T_{\text{fly}} \) is the single flight time of UAV drones; \( T_{\text{chg}} \) is the single battery更换 time of UAV drones. The effective inspection time \( T_{\text{eff}} \) can be approximated as:
$$ T_{\text{eff}} = T_{\text{fly}} – T_{\text{overhead}} $$
where \( T_{\text{overhead}} \) includes time for acceleration, deceleration, and turning. For simplicity, we can assume \( T_{\text{eff}} \approx T_{\text{fly}} \) if the overhead is negligible compared to the flight time.
The safety indicator \( R \) can be expressed as the reciprocal of the UAV drones damage probability, i.e., the probability that UAV drones complete the task without damage:
$$ R = 1 – P_d $$
where \( P_d \) is the damage probability of UAV drones in a single task. The damage probability \( P_d \) may depend on factors such as wind speed, obstacle density, and pilot skill. In mountainous areas, sudden wind gusts or proximity to cliffs can increase \( P_d \).
Combining inspection efficiency and safety, we can define an evaluation indicator \( I \), which incorporates considerations of both efficiency and safety:
$$ I = E \times R^{\beta} $$
where \( \beta \) is a weighting exponent that determines the relative importance of safety versus efficiency. If \( \beta = 1 \), efficiency and safety are weighted equally; if \( \beta > 1 \), safety is prioritized; if \( \beta < 1 \), efficiency is prioritized. This comprehensive indicator \( I \) can help in making trade-off choices between different partitions and different collaboration methods to achieve both efficient and safe inspection effects. The following table is an example created based on the above model.
| Partition ID | Partition Size S (m²) | UAV Drones Speed v (m/s) | Flight Time T_fly (min) | Battery Change Time T_chg (min) | Damage Probability P_d | Inspection Efficiency E (m²/min) | Safety Indicator R | Comprehensive Evaluation I (β=1) |
|---|---|---|---|---|---|---|---|---|
| 1 | 10000 | 5 | 10 | 5 | 0.01 | 1000 | 0.99 | 990 |
| 2 | 20000 | 5 | 15 | 5 | 0.02 | 1333.33 | 0.98 | 1306.67 |
| 3 | 15000 | 5 | 12 | 4 | 0.015 | 1250 | 0.985 | 1231.25 |
Table 3 displays the UAV drones inspection efficiency, safety indicators, and comprehensive evaluation for three partitions. The comprehensive evaluation \( I \) considers the balance between inspection efficiency and UAV drones safety. A larger value indicates better performance in both efficiency and safety for the inspection task in that partition. Through Table 3, the inspection performance of different partitions can be intuitively compared, providing a basis for optimizing UAV drones inspection tasks.
The vehicle-drone collaboration model can be extended to include multiple UAV drones and multiple ground vehicles. In such scenarios, task allocation and coordination become critical. We can formulate a multi-agent system where each UAV drones is assigned a subset of partitions, and ground vehicles are stationed at strategic locations for battery swapping or emergency response. The objective might be to minimize the total inspection time while ensuring that each UAV drones does not exceed its endurance limit. This can be modeled as a vehicle routing problem with time windows (VRPTW) or a capacitated vehicle routing problem (CVRP).
Let \( K \) be the set of UAV drones, and \( V \) be the set of ground vehicles. Each UAV drones \( k \in K \) has a battery capacity \( B_k \) and a maximum flight time \( T_{\text{max},k} \). Each partition \( i \) requires an inspection time \( t_i \) (which depends on the partition size and UAV drones speed). The goal is to assign partitions to UAV drones and schedule ground vehicle visits for battery swaps such that all partitions are inspected, and each UAV drones returns to base before its battery depletes. This can be expressed as a mixed-integer linear programming (MILP) problem:
$$ \min \sum_{k \in K} \sum_{i \in P} \sum_{j \in P} c_{ij} x_{ijk} $$
subject to:
$$ \sum_{k \in K} \sum_{j \in P} x_{ijk} = 1 \quad \forall i \in P $$
$$ \sum_{i \in P} x_{ihk} – \sum_{j \in P} x_{hjk} = 0 \quad \forall h \in P, \forall k \in K $$
$$ \sum_{j \in P} x_{0jk} = 1 \quad \forall k \in K $$
$$ \sum_{i \in P} x_{i0k} = 1 \quad \forall k \in K $$
$$ t_i + \sum_{j \in P} c_{ij} x_{ijk} \leq T_{\text{max},k} \quad \forall k \in K $$
where \( x_{ijk} \) is a binary variable indicating whether UAV drones \( k \) travels from partition \( i \) to partition \( j \), \( c_{ij} \) is the travel time between partitions, and index 0 represents the base station. This formulation ensures that each partition is visited exactly once, flow conservation for each UAV drones, and that each UAV drones starts and ends at the base, with total time not exceeding the maximum flight time.
Advanced Considerations for UAV Drones Path Planning
Beyond the core modules, several advanced factors can influence the performance of UAV drones inspections. These include environmental conditions, sensor capabilities, and data processing requirements. For instance, wind speed and direction can affect the stability and energy consumption of UAV drones. In mountainous regions, wind patterns can be unpredictable, necessitating adaptive path planning that adjusts in real-time. We can incorporate wind models into the path planning algorithm. Let \( \vec{w}(x,y) \) be the wind vector at location \( (x,y) \). The ground speed of UAV drones \( \vec{v}_g \) is related to its airspeed \( \vec{v}_a \) by:
$$ \vec{v}_g = \vec{v}_a + \vec{w} $$
To maintain a desired ground speed \( \vec{v}_d \), the UAV drones must adjust its airspeed accordingly. This adjustment may affect energy consumption, as flying against the wind requires more power. The energy consumption rate \( P \) can be modeled as:
$$ P = \frac{1}{2} \rho C_D A \|\vec{v}_a\|^3 + mg \cdot \frac{v_z}{\eta} $$
where \( \rho \) is air density, \( C_D \) is drag coefficient, \( A \) is cross-sectional area, \( m \) is mass, \( g \) is gravity, \( v_z \) is vertical speed, and \( \eta \) is propulsion efficiency. Integrating this over the flight path gives total energy consumption, which must be within battery capacity.
Another consideration is the resolution of imagery required for defect detection. The camera on UAV drones must capture images with sufficient resolution to identify cracks, hotspots, or soiling on photovoltaic panels. The ground sampling distance (GSD) is a key metric, defined as the distance between pixel centers measured on the ground. It is given by:
$$ \text{GSD} = \frac{h \cdot s}{f} $$
where \( h \) is flight altitude, \( s \) is sensor pixel size, and \( f \) is focal length. For example, to detect cracks smaller than 1 mm, the GSD must be less than 0.5 mm. This imposes constraints on flight altitude: lower altitude yields better resolution but reduces coverage per image, potentially increasing path length. Therefore, path planning must balance coverage efficiency with inspection quality.
Data processing and analysis are also crucial. UAV drones can generate vast amounts of imagery, which must be analyzed automatically to identify anomalies. Machine learning algorithms, such as convolutional neural networks (CNNs), can be trained to detect defects. The path planning can be optimized to ensure that images are captured with consistent lighting and角度 to facilitate automated analysis. For instance, avoiding backlight situations or ensuring the sun is at a favorable angle can improve image quality.
Simulation and Validation of UAV Drones Path Planning
To validate the proposed path planning framework, simulation studies can be conducted. Using software like MATLAB or ROS/Gazebo, we can create a digital twin of the photovoltaic power station and simulate UAV drones flights. The simulation can incorporate terrain models, photovoltaic string layouts, and environmental conditions. Key performance metrics include total inspection time, energy consumption, coverage percentage, and collision avoidance.
We can define a coverage ratio \( C_r \) as the proportion of the total area that is effectively inspected:
$$ C_r = \frac{A_{\text{covered}}}{A_{\text{total}}} $$
where \( A_{\text{covered}} \) is the area covered by the camera’s field of view during the flight, and \( A_{\text{total}} \) is the total area of all photovoltaic strings. Ideally, \( C_r \) should be 1, but in practice, some overlap or gaps may occur. The path planning algorithm should aim to maximize \( C_r \) while minimizing path length.
Collision avoidance is critical, especially in mountainous terrain with obstacles like poles, wires, or trees. We can use sensors like LiDAR or ultrasonic sensors on UAV drones for real-time obstacle detection. The path planning algorithm can incorporate potential fields or rapidly-exploring random trees (RRT) for dynamic obstacle avoidance. The overall path \( \mathbf{p}(t) \) can be planned using a combination of global planning (e.g., waypoints from genetic algorithm) and local planning (e.g., avoiding immediate obstacles).
Simulation results for the case power station might show that the proposed method reduces inspection time by 40% compared to manual inspection, while maintaining a coverage ratio above 99%. Energy consumption per inspection might be reduced by 30% through optimized paths. These improvements highlight the value of advanced path planning for UAV drones.
Conclusion
Through practical verification, this technical scheme not only significantly improves inspection efficiency but also ensures high accuracy of inspection results, fully demonstrating its practical value in the operation and maintenance management of mountainous centralized photovoltaic power stations. The proposal of this innovative technical route provides new ideas for the intelligent and efficient operation and maintenance management of photovoltaic power stations. The integration of area division, optimized path planning, and vehicle-drone collaboration creates a robust framework for UAV drones inspections. As UAV drones technology continues to advance, with improvements in battery life, autonomy, and sensor capabilities, the proposed methods will become even more effective. Future work may focus on real-time adaptive planning, multi-UAV drones swarms, and integration with artificial intelligence for predictive maintenance. The widespread adoption of UAV drones in photovoltaic inspections will contribute to the reliability and sustainability of solar energy generation, supporting the global transition to renewable energy.
In summary, this paper has presented a comprehensive approach to UAV drones path planning for mountainous centralized photovoltaic power stations. By leveraging algorithms like DBSCAN, genetic algorithms, and collaborative models, we can enhance the efficiency, safety, and accuracy of inspections. The mathematical models and tables provided offer practical tools for implementation. As the renewable energy sector grows, optimized UAV drones operations will play an increasingly vital role in ensuring the performance and longevity of photovoltaic installations.