In addressing the challenge of load restoration in large-scale information blind spots caused by communication interruptions that delay fault recovery, I propose a fault recovery strategy for active distribution networks that considers the restoration of temporary communication using unmanned aerial vehicles (UAVs), also known as drone technology. This strategy aims to maximize the amount of fault load recovery by coordinating dispatchable resources within the region and synchronously implementing island partitioning and network topology reconfiguration. This involves the collaborative optimization of drone deployment schemes and emergency repair scheduling.
Introduction
The high proportion integration of distributed generation and electric vehicles has significantly altered the physical structure and power flow distribution of distribution networks. Under normal operating conditions, the strong coupling between information systems and physical systems ensures the safe and reliable operation of the power grid. However, during extreme disasters, this deep coupling characteristic can increase system vulnerability, leading to more complex problems for the distribution system. Traditional recovery strategies for distribution networks are no longer sufficient to meet current development needs. The critical challenge for future research is constructing efficient and coordinated recovery plans when facing coupled failures of communication and physical systems.
In extreme natural disasters causing large-scale communication system paralysis, emergency communication technology, particularly drone technology, can effectively enhance the rapid self-healing capability of the communication network. The application of drone technology in emergency communication has attracted significant attention. Some studies have explored methods that integrate vehicle-to-grid (V2G) technology with drone communication for collaborative recovery, using drones to rebuild regional communication networks after a disaster while forming island power supplies through V2G. Other research has focused on the drone-assisted device-to-device relay communication reconstruction technology for emergency communication needs. Studies on drone deployment optimization have solved for the optimal spatial configuration of drones. Investigations have also delved into the correlation mechanism between communication system vulnerability and power load restoration. For power system fault diagnosis, research has systematically addressed drone deployment strategies during the disaster warning phase and path optimization for post-disaster inspections. Some work has developed transmission network restoration methods based on co-evolutionary computation, significantly improving restoration efficiency. Frameworks considering communication network constraints in optimal power flow calculations, using multi-stage bi-level optimization models, have been constructed for collaborative restoration of power cyber-physical systems. From the perspective of information system optimization, improving communication routing algorithms has effectively enhanced the disaster resilience of power cyber-physical systems. Despite these advances, research on using emergency communication devices like drones for efficient communication reconstruction in fault scenarios of distribution network cyber-physical systems remains relatively scarce. Furthermore, the collaborative optimization between rapid communication network reconfiguration and timely power network restoration requires more in-depth investigation.
My work focuses on the problem of load restoration in large-scale information blind spots, which causes delays in fault recovery due to communication interruptions. I propose a fault recovery strategy for active distribution networks that leverages drone technology to restore temporary communication. This strategy coordinates dispatchable resources within the region, implements island partitioning and network topology reconfiguration simultaneously, and achieves collaborative optimization between drone deployment and repair scheduling to maximize the recovery of fault loads.
Communication Network Fast Recovery Model Considering Drone Technology
Drone Communication Model
During the deployment and path planning of drone emergency communication networks, the fundamental spatial constraints must be satisfied. In my model, after an extreme disaster, a drone receives instructions to depart from an emergency warehouse or a candidate stopping point i to travel to candidate stopping point j to establish a temporary communication network. The flight time $$t_{i j}^{UAV}$$ is calculated along the pre-planned optimal path representing the shortest time between the two nodes. The operation time for setting up the emergency communication network is $$t_{op, j}^{UAV}$$, and the hovering time in the air is $$t_{stay, j}^{UAV}$$. The drone’s maximum endurance time during the fault recovery period is $$t_{C, max}^{UAV}$$. If the drone departs from the emergency warehouse at time T0 and returns to it at time T1, the constraint is $$(T_{1} – T_{0}) \leq t_{C, max}^{UAV}$$.
The basic spatial constraints in the model are: a drone can travel round-trip from the emergency warehouse to a candidate stopping point; a drone can only travel one-way between candidate stopping points (non-emergency warehouse points); and a drone proceeds to the next area for communication restoration only after all fault lines in the area it covers have been repaired. These constraints are represented by the following equations in my optimization model:
$$x_{z, j, u} + x_{j, z, u} \leq 2$$
$$x_{i, j, u} + x_{j, i, u} \leq 1$$
$$x_{j, K, u} \left[ 1 – y_{j, k, u} \sum_{mn \in \zeta} (1 – G_{mn}) \right] = 1$$
In these equations, z is the emergency warehouse point, and $$x_{z, j, u}$$ is a binary variable indicating whether drone u departs from emergency warehouse point z to candidate stopping point j. Similarly, $$x_{i, j, u}$$ is a binary variable for drone u traveling from candidate stopping point i to candidate stopping point j. The set of faulty physical lines is denoted as $$\zeta$$. The binary variable $$x_{j, K, u}$$ indicates whether drone u proceeds to candidate stopping point j to deploy the drone emergency communication mode K. $$G_{mn}$$ is a binary variable representing whether the physical fault line mn has been repaired. Finally, $$y_{j, k, u}$$ is a binary variable indicating whether the FTU device in the information blind spot is covered by the communication from drone u at candidate stopping point j using drone emergency communication mode K.
Phased Coverage Model Considering Drone Technology
The target area is divided into multiple sets of information blind spots. For each blind spot, the value of the load controlled by the FTU device within it is considered. The deployment of drone-assisted emergency communication aims to maximize the total value of loads controlled by the covered information blind spots. I have improved the maximum coverage model to formulate the objective function and constraints for this drone technology application:
$$\max f = \sum_{d \in D} p_{d} z_{d}$$
Subject to:
$$\sum_{j \in J_{UAV}} x_{jK} \leq N_{UAV}$$
$$z_{d} = \begin{cases} 1, & \sum_{a \in A_{d}} (1 – y_{d a j K}) = 0 \\ 0, & \sum_{a \in A_{d}} (1 – y_{d a j K}) \geq 1 \end{cases}$$
$$\sum_{j \in J_{UAV}} a_{d j} x_{jK} \geq N_{d} z_{d}, \forall d \in D$$
In this formulation, D is the set of information blind spots. The binary variable $$z_{d}$$ indicates whether blind spot d is fully covered. $$x_{jK}$$ is a binary variable for deploying drone emergency communication mode K at candidate stopping point j. $$N_{UAV}$$ is the number of drones available. The binary variable $$y_{d a j K}$$ indicates whether FTU device a in blind spot d is covered by drone communication at candidate point j. The binary variable $$a_{d j}$$ indicates whether a drone at candidate point j can cover the FTU devices in blind spot d. $$N_{d}$$ is the minimum number of emergency devices needed to fully cover blind spot d.
The priority weight for blind spot coverage is defined based on the value hd and urgency v of the load controlled by the FTU device, calculated as $$p_{d} = h_{d} v$$, and sorted in descending order of $$p_{d}$$. The communication range constraint must also be satisfied, which involves calculating the Euclidean distance between the FTU device location and the drone location.
Active Distribution Network Fault Recovery Model
Objective Function
After a disaster, my primary goal is to minimize the weighted power loss over the recovery time by fully restoring power to lost loads through island partitioning and network reconfiguration, while simultaneously carrying out the repair of damaged lines. The objective function for this stage is defined as:
$$\min f = c_{1} \sum_{i=1}^{n} a_{i} w_{i} P_{i,t} \Delta t + \mu \Delta t_{n”’}^{a}$$
In this function, $$c_{1}$$ is the cost coefficient for lost load. $$w_{i}$$ is the importance weight for load node i. $$a_{i}$$ is a binary variable indicating whether node i is supplied with power. $$P_{i,t}$$ is the active power consumed by node i at time interval t. $$\mu$$ is the weight coefficient for repair time. $$\Delta t$$ is the length of the time interval. $$\Delta t_{n”’}^{a}$$ represents the sum of travel time and repair time for the simulated repair of the $$n”’$$ faulty line.
After full restoration of power to all lost loads, the scheduling of the remaining repairs is determined to minimize the total time:
$$f = \min \left\{ \max T_{sa} \right\}$$
Here, $$T_{sa}$$ is the total time required for the physical repair team sa to complete the remaining repair tasks, which is the sum of the travel time $$T_{move, sa}$$ and the line repair time $$T_{mn, rep, sa}$$: $$T_{sa} = T_{move, sa} + T_{mn, rep, sa}$$.
Constraints
The constraints in my fault recovery model include:
- Load reduction constraints: Some loads are contracted for permissible reduction during peak periods or faults.
- Energy storage constraints: Storage devices are used to manage the volatility of renewable sources and support planned island power supply during fault recovery.
- Power balance constraints: After recovery, power balance must be maintained at each node within the distribution network subsystems.
- Branch power flow constraints: The power flow through each branch must be within its limits: $$P_{ij} \leq P_{ij, max}$$.
- Node voltage constraints: The voltage at each node must be within acceptable limits: $$U_{i, min} \leq U_{i} \leq U_{i, max}$$.
- Physical repair team constraints: These include resource constraints for task allocation and path planning, and time constraints for task transfer and line repair status.
For the repair team constraints, key equations are:
$$\sum_{sa=1}^{S} k_{mn, sa} = 1, \forall mn \in \zeta$$
$$\sum_{sa=1}^{S} k_{mn, sa} g_{mn \rightarrow kl, sa} = 1, \forall mn, kl \in \zeta$$
In these, $$k_{mn, sa}$$ is a binary variable indicating if repair team sa is responsible for fault line mn. $$g_{mn \rightarrow kl, sa}$$ indicates if the repair team sa repairs line mn before line kl. The time constraint is:
$$ \begin{cases} T_{mn \rightarrow kl, end, sa} = T_{mn, start, sa} + t_{mn, rep, sa} + t_{move, sa} \\ T_{kl, start, sa} = T_{mn \rightarrow kl, end, sa} \end{cases}, \forall mn, kl \in \zeta $$
$$ G_{mn} = \begin{cases} 0, & T_{now} < T_{mn \rightarrow kl, end, sa} \\ 1, & T_{now} \geq T_{mn \rightarrow kl, end, sa} \end{cases}, \forall mn \in \zeta $$
Steps for the Proposed Fault Recovery Strategy
The process for implementing my proposed fault recovery strategy is as follows:
- Data reading and initialization: Read distribution network parameters, information on grid-connected photovoltaic-storage and wind-storage nodes and their output, emergency warehouse locations, number of repair teams and drones, and input fault data.
- Fault isolation and initial recovery: Identify the de-energized load area and information blind spots. In non-information blind areas, prioritize the use of distributed generation with black-start capability to form islands for load restoration. Use network reconfiguration for parts that cannot operate as islands. Jointly optimize island partitioning and reconfiguration to obtain a staged optimal solution.
- Phased drone communication coverage: Determine the recovery priority based on the importance of the load controlled by FTUs in the blind spot. Considering the blind spot scope and equipment arrival time, determine drone stopping points and quantities. Deploy drones to establish a temporary communication network under time and spatial constraints. Once blind spot communication is restored, the regional power supply can form an island to restore load.
- Re-dispatch after communication restoration: After communication for some blind spots is restored, dispatch power supplies with island operation capability. Continue network reconfiguration optimization for the unrecovered areas to achieve broader power supply restoration.
- Repair priority optimization: Before each repair, simulate the load recovery after repairing different lines. Quantify and compare to determine the repair sequence. After each recovery, update the system state and dynamically adjust the plan. When lines in a blind spot are repaired and circuit breakers can close, dispatch a drone to the next blind spot.
- Check for full restoration: If all de-energized loads are restored, arrange the repair teams to minimize total repair time for the remaining physical lines. Otherwise, return to step 5.
- Completion: The post-disaster recovery ends when all faulty lines are repaired.
Case Study
Case Setup
To validate the effectiveness of my proposed fault recovery strategy, I used an improved IEEE 33-node distribution system with its supporting communication network as an example. The fault scenario after an extreme natural disaster involves both communication and physical layers. Key parameters are set as follows:
| Parameter | Value |
|---|---|
| Emergency warehouse location | Node 12 |
| Cost coefficient for lost load | 5 yuan/(kW·h) |
| Distributed generation nodes | 13 (wind-storage), 17 (PV-storage), 21 (PV-storage), 24 (PV-storage), 30 (wind-storage) |
| Energy storage unit capacity | 100 kW·h |
| Energy storage rated power | 20 kW |
| Disaster occurrence time | 09:30 AM |
| Number of physical repair teams | 2 |
| Fault repair time per location | 40 minutes |
| Drone flight speed | 80 km/h |
| Drone communication coverage radius | 3 km |
| Number of primary drones | 3 |
| Candidate stopping nodes | 5, 6, 17, 20, 21, 24, 25, 30 |
| Physical network faults (lines) | 6 (4-5, 8-9, 15-16, 20-21, 23-24, 26-27) |
| Communication network faults (nodes) | 4 (nodes 2, 3, 4, 8, with backup power) |
I assume that all physical nodes are without backup power. To maximize power supply restoration, distributed generation with black-start capability will form planned islands with some de-energized loads.
Fault Recovery Result Analysis
For efficient power supply restoration, I first collected the status information of the distribution network’s physical and communication systems. I used a joint optimization strategy of island partitioning and network reconfiguration for fault handling. Initially, after the communication and physical systems are both damaged and before any repair work begins, the initial recovery topology is established.
Fast Communication Network Recovery Using Drone Technology
First, I dispatched drones to target areas to build temporary communication networks. The deployment of drone technology is summarized in the following table:
| Drone ID | Starting Point | Destination Stopping Point | Travel Time (min) |
|---|---|---|---|
| 1 | 12 | 5 | 8.7 |
| 2 | 12 | 21 | 9.3 |
| 3 | 12 | 24 | 9.6 |
| 1 (re-tasked) | 24 | 17 | 13.7 |
Using a graded and zoned guarantee strategy, drones were dispatched to stopping points 5, 21, and 24. With a given drone deployment time of 5 minutes, I calculated the total time required for each drone to establish a complete emergency communication network.
Fault Recovery After Partial Communication Restoration
Using the coordinated optimization strategy of island partitioning and network reconfiguration, I obtained the optimal recovery plan for the current stage.
- 09:30 – Communication Established: Temporary communication networks were established by drones at points 5, 21, and 24.
- 11:00 – Line 4-5 Repaired: The repair of line 4-5 allowed for the restoration of the important load at node 5.
- 11:08 – Line 23-24 Repaired: With the repair of line 23-24, the island in area 2 was able to connect to the grid.
- 11:30 – Drone Re-deployed: Drone 1 established a temporary communication network at stopping point 17, guiding the restoration of load at nodes 15, 16, 17, and 18.
This process led to the successful recovery of loads at nodes 5, 15, 16, 17, 18, and 26.
Scheme Comparison and Results
I compared my proposed fault recovery strategy (Scheme 2), which leverages drone technology for emergency communication restoration, with a strategy that relies on manual restoration of the communication network (Scheme 1). The comparative results for power loss recovery over time are presented in the table below.
| Time | Scheme 1 Power Loss Recovery (kW) | Scheme 2 Power Loss Recovery (kW) |
|---|---|---|
| 09:30 | 0 | 0 |
| 10:00 | 200 | 900 |
| 10:30 | 450 | 1800 |
| 11:00 | 750 | 2100 |
| 11:30 | 1000 | 2900 |
| 12:00 | 1400 | 3200 |
| 12:30 | 1600 | 3200 |
| 13:00 | 2000 | 3200 |
| 13:30 | 2400 | 3200 |
| 14:00 | 2800 | 3200 |
| 14:30 | 3200 | 3200 |
| 15:00 | 3200 | 3200 |
The results clearly demonstrate the effectiveness of my strategy. Scheme 2, which integrates drone technology for communication recovery, achieves full power restoration significantly faster than Scheme 1. While Scheme 1 takes approximately 5 hours from the onset of the disaster at 09:30 to fully recover by 14:30, Scheme 2 achieves complete recovery in just 2 hours, by 11:30.
The total time for complete power failure recovery in my proposed strategy is reduced by 60% compared to the manual communication network repair strategy. The key difference lies in the ability of drone technology to rapidly establish temporary communication networks. This allows for the immediate coordination of dispatchable resources, enabling the formation of stable islands and the optimization of network topology in the information blind spots. This accelerates the deployment of repair resources and minimizes the duration and impact of power outages.
Conclusion
In conclusion, my proposed fault recovery strategy for active distribution networks, which leverages drone technology to restore temporary communication, offers a significant advancement in addressing the challenge of load restoration in large-scale information blind spots. The simulation results verified the following:
- Maximized short-term load recovery: By first restoring temporary communication with drones and then using a joint optimization strategy of island partitioning and network reconfiguration, the amount of load recovered in a short time is maximized.
- Significant time savings: Compared to a strategy relying on manual communication network repair, my proposed strategy using drone technology reduces the total time for full power load recovery by 60%. It also achieves a higher load recovery rate during the same recovery period.
The integration of drone technology is central to this improvement. It provides a rapid and flexible means to bridge communication gaps, enabling smarter, faster, and more effective coordination of distributed energy resources and repair efforts, thereby enhancing the resilience of active distribution networks against extreme disasters.