| CPC G05D 1/644 (2024.01) [G05D 1/689 (2024.01); G05D 1/69 (2024.01); H04N 7/181 (2013.01); G05D 2105/89 (2024.01); G05D 2109/254 (2024.01)] | 11 Claims |
|
1. A method for stochastic inspections on power grid lines based on unmanned aerial vehicle-assisted edge computing, wherein an inspection is conducted on a target power gird area including power grid equipment and power transmission lines by applying an unmanned aerial vehicle group including M inspection unmanned aerial vehicles and a superior unmanned aerial vehicle based on a central base station arranged on a fixed position; comprising following steps:
Step S1, constructing, based on a flight mode of each of the inspection unmanned aerial vehicles in the unmanned aerial vehicle group, an unmanned aerial vehicle-assisted power grid lines stochastic inspection system, wherein the inspection unmanned aerial vehicles are merely in charge of acquiring video images for the power gird equipment and the power transmission lines in the target power gird area, and data are processed on obtained video images by the superior unmanned aerial vehicle or the central base station, and then entering Step S2;
Step S2, acquiring, by each of the inspection unmanned aerial vehicles in the unmanned aerial vehicle group, the video images for the power gird equipment and the power transmission lines in the target power gird area based on the unmanned aerial vehicle-assisted power grid lines stochastic inspection system, and obtaining the video image data acquired and obtained by each of the inspection unmanned aerial vehicles corresponding to each time slot respectively, and then entering Step S3;
Step S3, constructing, according to the video image data acquired and obtained by each of the inspection unmanned aerial vehicles corresponding to each time slot respectively, a digital twin network of the unmanned aerial vehicle-assisted power grid lines stochastic inspection system, in combination with a weight, a signal transmission power and position coordinates of each of the inspection unmanned aerial vehicles, a weight, a signal transmission power, position coordinates, and a computing capacity of the superior unmanned aerial vehicle, position coordinates of the central base station, as well as a system communication bandwidth, to fit the position coordinates of each of the inspection unmanned aerial vehicles and the superior unmanned aerial vehicle, and a resource status of the system, and then entering Step S4;
Step S4, constructing, based on constraints of an offload latency and a data task processing latency for the power grid lines stochastic inspection system, an energy consumption model or a balanced energy consumption model of the unmanned aerial vehicle group corresponding to each time slot respectively, according to the digital twin network of the unmanned aerial vehicle-assisted power grid lines stochastic inspection system; further constructing an objective function for minimizing energy consumption of the unmanned aerial vehicle group corresponding to each time slot respectively or an objective function for minimizing balanced energy consumption of the unmanned aerial vehicle group corresponding to each time slot respectively, and then entering Step S5;
Step S5, randomly initializing the position coordinates of the superior unmanned aerial vehicle, constructing, based on the position coordinates and the video image data of each of the inspection unmanned aerial vehicles corresponding to a t-th time slot respectively, a system status at the t-th time slot, and then entering Step S6;
Step S6, solving, by adopting a deep deterministic policy gradient algorithm in a deep reinforcement learning, the energy consumption model of the unmanned aerial vehicle group corresponding to each time slot respectively, based on the position coordinates of the superior unmanned aerial vehicle and the system status at the t-th time slot, according to the objective function for minimizing energy consumption of the unmanned aerial vehicle group corresponding to each time slot or the objective function for minimizing balanced energy consumption of the unmanned aerial vehicle group corresponding to each time slot respectively; obtaining, an action space of the system at the t-th time slot corresponding to the system status at the t-th time slot in combination with the position coordinates of the superior unmanned aerial vehicle, that is, the action space of the system at the t-th time slot corresponding to the system status at the t-th time slot in combination with the position coordinates of the superior unmanned aerial vehicle, wherein the action space of the system at the t-th time slot is composed of the signal transmission power of each of the inspection unmanned aerial vehicles corresponding to the t-th time slot respectively, an offload mode of each of the inspection unmanned aerial vehicles corresponding to the t-th time slot respectively regarding the superior unmanned aerial vehicle or the central base station, and the signal transmission power and an allocated CPU calculation frequency of the superior unmanned aerial vehicle corresponding to the t-th time slot, and then entering Step S7;
Step S7, determining whether an iteration overflow condition is satisfied or not, if yes, entering Step S8, if no, solving and updating, by using a genetic algorithm, the position coordinates of the superior unmanned aerial vehicle, based on the system status at the t-th time slot, in combination with system resource allocations and offload decision schemes for the video image data in the action space of the system at the t-th time slot corresponding to the position coordinates of the superior unmanned aerial vehicle, and returning to Step S6; and
Step S8, processing, according to the position coordinates of the superior unmanned aerial vehicle, and the system resource allocations and the offload decision schemes for the video image data in the action space of the corresponding system at the t-th time slot, the video image acquired by each of the inspection unmanned aerial vehicles corresponding to each time slot in Step S2, to offload the video image data to the superior unmanned aerial vehicle or the central base station for processing.
|