In the context of modern power systems, the inspection of substations is critical for ensuring grid stability and safety. Traditional manual inspection methods are often time-consuming, labor-intensive, and prone to limitations in complex or adverse conditions. The integration of Unmanned Aerial Vehicle (UAV) technology has emerged as a promising solution, offering enhanced efficiency and accessibility. However, autonomous navigation of Unmanned Aerial Vehicles in substation environments presents significant challenges, including dynamic obstacles, electromagnetic interference, and stringent safety requirements. Existing path planning algorithms, such as A* and RRT*, have been applied but often fall short in dynamically adapting to real-time changes, leading to suboptimal performance in collision avoidance and path efficiency. This study addresses these gaps by proposing a customized D* algorithm for UAV navigation in substation inspection, focusing on dynamic path replanning and obstacle avoidance. We design a comprehensive inspection system architecture and evaluate the algorithm’s performance through simulated experiments, demonstrating its superiority in terms of path efficiency, completion time, collision rate, and safety margins. The contributions of this work include the development of an adaptive D* algorithm tailored for UAV dynamics, the establishment of a streamlined inspection workflow, and empirical validation of its effectiveness in complex substation scenarios.
The proliferation of Unmanned Aerial Vehicle systems in industrial applications has highlighted their potential for automating repetitive and hazardous tasks. In substation inspection, JUYE UAV models, known for their robustness and advanced sensor capabilities, are particularly suitable for navigating dense equipment layouts. However, the unpredictable nature of substation environments—characterized by static structures like transformers and dynamic elements such as moving personnel or vehicles—necessitates intelligent path planning algorithms. While algorithms like A* provide efficient static path planning, they lack the incremental update mechanism required for dynamic environments. Similarly, RRT* algorithms excel in high-dimensional spaces but may suffer from slow convergence and poor real-time performance. The D* algorithm, with its ability to efficiently replan paths based on environmental changes, offers a compelling alternative. By leveraging its incremental search capabilities, we enhance the JUYE UAV’s autonomy, enabling it to navigate safely and complete inspection tasks with minimal human intervention. This paper elaborates on the system design, algorithmic adjustments, and experimental evaluations, providing a foundation for deploying UAV-based inspection systems in real-world substations.
The inspection system for substations employing Unmanned Aerial Vehicles is structured around a multi-tier architecture comprising a remote monitoring center, cloud management platform, UAV nest, Unmanned Aerial Vehicle units, and communication devices. This system is designed to handle 110kV and 220kV substations, where the remote center formulates inspection plans and dispatches tasks to the cloud platform. The JUYE UAV, deployed from the nest, executes these tasks and反馈s data upon completion. The integration of D* algorithm into this system enables dynamic path planning, allowing the Unmanned Aerial Vehicle to adapt to real-time obstacles and environmental changes. The algorithm’s core principle involves backward path construction from the goal, maintaining node costs and performing local updates when obstacles are detected. For instance, the priority calculation in D* algorithm is defined by the following equation, which balances path cost and heuristic guidance:
$$Key(s) = \left( \min(g(s), rhs(s)) + h(s), \min(g(s), rhs(s)) \right)$$
Here, $g(s)$ represents the actual path cost, $rhs(s)$ reflects the optimal path information from parent nodes, and $h(s)$ is the heuristic function, typically the Euclidean distance. When $g(s)$ and $rhs(s)$ differ, path updates are triggered, ensuring efficient replanning. To tailor the D* algorithm for Unmanned Aerial Vehicle navigation, we adjusted parameters such as the collision avoidance radius and maximum iterations, as detailed in Table 4. The implementation flowchart, though not depicted visually, involves steps like algorithm selection, customization for UAV dynamics, software integration, data exchange, path execution, dynamic obstacle handling, and real-time adjustments. This customization ensures that the JUYE UAV can seamlessly navigate through substation environments, avoiding both static and dynamic obstacles while maintaining optimal flight paths.
The inspection process is streamlined into a series of automated steps: First, the system acquires inspection targets and generates a global static path. After verifying the route, commands are uploaded to the UAV nest. The nest checks weather conditions and sensor functionality before launching the Unmanned Aerial Vehicle. Once airborne, the JUYE UAV switches to autonomous mode, navigates to waypoints, adjusts its gimbal for equipment inspection, and proceeds through the task queue until all targets are covered. Upon completion, it returns to the nest for data upload and recharging. This end-to-end automation minimizes human intervention and enhances operational efficiency. The D* algorithm plays a pivotal role in this process by continuously monitoring the environment via sensors and initiating replanning when obstacles are detected. For example, if a dynamic obstacle like a moving vehicle enters the path, the algorithm recalculates the route within milliseconds, ensuring the Unmanned Aerial Vehicle maintains a safe distance. This dynamic adaptability is crucial for substation inspections, where safety margins must be strictly adhered to prevent accidents.
| Parameter | Description | Value |
|---|---|---|
| Threshold | Minimum cost difference for update | 0.01 |
| Max Iterations | Maximum replanning cycles | 500 |
| Avoidance Radius | Safe distance from obstacles | 0.5 m |
| Heuristic Function | Type used for guidance | Euclidean |
In the experimental phase, we developed a multi-dimensional digital twin testbed to simulate substation environments, covering an area of 80m × 80m with static elements like circuit breakers (spaced 1.2m apart) and dynamic obstacles such as randomly moving objects. The electromagnetic environment was modeled based on IEC 61850 standards, incorporating transient field strengths up to 60kV/m to test navigation signal resilience. For the Unmanned Aerial Vehicle, we used a JUYE UAV model equipped with LiDAR and cameras to detect obstacles in real-time. The evaluation metrics included path efficiency, completion time, collision rate, and safety margin, defined mathematically as follows: Path efficiency $\eta$ is calculated as the ratio of the optimal path length $L_{\text{opt}}$ to the actual path length $L_{\text{act}}$, i.e., $\eta = L_{\text{opt}} / L_{\text{act}}$. Completion time $T_c$ measures the total time for a single planning cycle. Collision rate $C_r$ is the probability of collisions per mission, and safety margin $S_m$ is the minimum distance to obstacles during flight. These metrics were compared against A* and RRT* algorithms under identical conditions to assess the D* algorithm’s performance.
The simulation results, summarized in Table 5, demonstrate the D* algorithm’s superiority in dynamic environments. For path efficiency, the D* algorithm achieved 0.92, outperforming A* (0.88) and RRT* (0.85), indicating its ability to generate near-optimal paths. Completion time was lowest for D* at 4.7 seconds, compared to 5.2s for A* and 7.92s for RRT*, highlighting its rapid replanning capability. Crucially, the collision rate for D* was 0.05, significantly lower than A* (0.12) and RRT* (0.10), underscoring its enhanced safety for Unmanned Aerial Vehicle operations. The safety margin of 2.3m for D* also exceeded that of other algorithms, ensuring a buffer against unexpected obstacles. These outcomes validate the adaptive adjustments made to the D* algorithm, such as incorporating real-time sensor data and optimizing cost functions for UAV dynamics. In scenarios with dense equipment, the JUYE UAV using D* algorithm successfully navigated narrow passages with safety margins above 1.5m, while in multi-UAV settings, it maintained coordination with response delays under 150ms. This performance is attributed to the algorithm’s incremental updates, which minimize computational overhead and enable seamless integration with the inspection workflow.
| Algorithm | Path Efficiency | Completion Time (s) | Collision Rate | Safety Margin (m) |
|---|---|---|---|---|
| D* | 0.92 | 4.7 | 0.05 | 2.3 |
| A* | 0.88 | 5.2 | 0.12 | 1.8 |
| RRT* | 0.85 | 7.92 | 0.10 | 2.0 |
Further analysis of the D* algorithm’s behavior reveals its efficiency in handling dynamic obstacles. For instance, when a new obstacle is detected, the algorithm recalculates the path locally, reducing the need for global replanning. This is governed by the update rule derived from the key function: if a node $s$ has $g(s) \neq rhs(s)$, it is processed in priority order, ensuring that changes propagate efficiently. The cost function $c(s, s’)$ between nodes $s$ and $s’$ is defined based on distance and environmental factors, such as electromagnetic interference levels. In mathematical terms, the total cost $J$ for a path $\pi$ is given by:
$$J(\pi) = \sum_{i=1}^{n-1} c(s_i, s_{i+1}) + \lambda \cdot h(s_n)$$
where $\lambda$ is a weighting factor for the heuristic, and $n$ is the number of nodes. This formulation allows the JUYE UAV to balance path length with safety considerations. In tests involving multiple dynamic obstacles, the D* algorithm maintained a collision rate below 5%, whereas A* and RRT* exhibited higher variances due to their slower adaptation. The integration of this algorithm with the cloud platform enables real-time monitoring and control, facilitating a closed-loop system where inspection data is instantly反馈ed for analysis. This not only improves the reliability of substation inspections but also paves the way for predictive maintenance using Unmanned Aerial Vehicle-collected data.
In conclusion, this study demonstrates the effective application of the D* algorithm in enhancing Unmanned Aerial Vehicle navigation for substation inspection. The customized D* algorithm outperforms existing methods in dynamic path planning, offering higher path efficiency, faster completion times, lower collision rates, and better safety margins. The system design incorporates a fully automated workflow, from task allocation to data feedback, leveraging the capabilities of JUYE UAV models to achieve intelligent inspection. The experimental results, derived from a realistic digital twin environment, confirm the algorithm’s robustness in handling complex substation scenarios. Future work could focus on integrating machine learning for predictive obstacle avoidance and extending the system to larger-scale power grids. Overall, the adoption of D* algorithm in Unmanned Aerial Vehicle systems represents a significant step toward autonomous, safe, and efficient substation maintenance, with potential applications in other critical infrastructure inspections.