Artificial intelligence system and machine learning method for electrical grid management using unmanned aerial vehicle traffic data collection
View Patent ↗An artificial intelligence system including a machine learning module for assisting in electrical grid maintenance. The machine learning module receiving data sets from UAVs travelling in the proximity of power lines and communicating sensed observations to a data library for training the module to spot markers of potential hazards and calculating a probability of hazards occurring under various conditions. The system communicating hazards of a type or above a predetermined level of probability to a communication center or utility company to remediate or schedule preventative maintenance.
1 . A method for identifying a vegetation condition endangering a power line, the method comprising:
providing one or more data collectors on a UAV;
collecting data with said one or more data collectors while the UAV is travelling in proximity to the power line;
receiving data from the UAV;
receiving operational data from a Supervisory Control and Data Acquisition (SCADA) system associated with the power line, wherein said operational data includes records of electrical anomalies or faults;
identifying a vegetation related hazard likely to damage the power line by correlating the data collected by the UAV with said operational data from the SCADA system;
determining a probability of the vegetation related hazard occurring within a predefined amount of time based on extrapolating vegetation growth data from historical observations;
determining when said probability exceeds a predetermined threshold;
wherein said UAV is one of a plurality of UAVs that traveled in proximity to the power line across different geographic regions of an electrical grid, the method further comprising:
building a prediction model based on pooling historical data from visual and non-visual data collectors associated with the plurality of UAVs that traveled in proximity to the power line to train the prediction model with data reflecting varying environmental conditions; and
wherein the prediction model determines the probability of the vegetation related hazard occurring within said predefined amount of time.
2 . The method of claim 1 , further comprising: receiving data from the UAV comprises uploading the data from the UAV to a system comprising a data library.
3 . The method of claim 2 , wherein the one or more data collectors include a camera configured to capture still visual images.
4 . The method of claim 3 , wherein the data received from the UAV is real time image data.
5 . The method of claim 1 , wherein the UAV includes a locating component comprising at least one of a magnetic detector and a heat detector configured to detect electromagnetic signatures of the power line for navigation along the power line in coordination with a stored map of the electrical grid.
6 . The method of claim 1 , wherein said UAV is one of a plurality of UAVs that traveled in proximity to the power line, the method further comprising: storing a machine learning module on server, wherein the machine learning module builds the prediction model based on said historical data.
7 . The method of claim 1 , further comprising: sending a communication to a predetermined recipient, the communication including a location of the vegetation related hazard.
8 . The method of claim 7 , wherein the communication includes a photograph taken from an angle no greater than 15 degrees looking down toward the vegetation related hazard.
9 . A system configured to identify a vegetation condition endangering a power line, the system comprising:
a UAV configured to fly in proximity of the power line;
one or more data collectors on the UAV, wherein the UAV is configured to collect observational data from the power line during one or more flights using said one or more data collectors;
a server having a machine learning module for receiving said observational data from said UAV and receiving operational data from a SCADA system associated with the power line, and calculating a probability of a vegetation related hazard occurring within a predefined amount of time by correlating said observational data with said operational data and determining whether the probability is above a predetermined threshold;
a communications link from said server for communicating information about the vegetation related hazard to a predetermined recipient;
wherein the predetermined recipient is a utility company, the system further comprising:
a prediction model built by the machine learning module and stored on said server, the prediction model built by pooling observational data from a plurality of UAVs operating across different geographic regions of an electrical grid, said prediction model including historical data from said one or more data collectors; and
wherein the prediction model determines the probability of the vegetation related hazard occurring within said predefined amount of time.
10 . The system of claim 9 , wherein said information in the communication includes a photograph taken from an angle no greater than 15 degrees looking down toward the vegetation related hazard.
11 . The system of claim 10 , wherein said UAV is one of a plurality of UAVs that traveled in proximity to the power line collecting historical data with a respective plurality of data collectors.