Objectives of the service
The overall goal of this study is to create an end to end electric grid monitoring platform using satellite images combined with high-resolution drone data to create multi-layer situational awareness, especially in the case of extreme weather events in the electric grid. The infrastructure systems and communities relying on these systems are highly vulnerable to extreme weather events such as hurricanes, storms, floods, landslides, and wildfires which have a higher frequency and impact as a result of climate change.
Users and their needs
Transmission System Operators and Distribution System Operators share the responsibility of delivering stable electricity to consumers, the business and industry, and the public service sector. One of the greatest opportunities to increase the resilience, and lower the cost of distribution is by making the vegetation management more efficient. By adding data originating from satellites together with intelligence gathered from drone or helicopter images, and smart meter data, we have the potential to improve detection of risk areas and tracking vegetation growth related to the utility infrastructure .
The use of satellite images for the analysis and prediction of the influence of vegetation on the risk level in power grids is a high-priority research axis because vegetation is one of the main factors for grid disturbances. But the potential behind the use of satellite-based information is much broader and can for example also be exploited for subsidence detections along power line trails. This easily shows the value that can be provided by adding step by step more and more satellite-based data to current state-of-the-art risk analysis approaches. Such information, in combination with detection of asset defects on pylons based on drone imagery, can be used for proactive monitoring and control, in addition to significantly improve risk assessment and preventive actions.
Service/ system concept
As in any common industry and business, the management of power-grids suffers nowadays from silo-based approaches, which hinders the optimization of the analytics to be made regarding risk handling. The main benefit of GridEyeS is to break those barriers by bringing together cross-disciplinary experts and combining state of the art techniques in some of the most promising fields, such as dynamic weather predictions, artificial intelligence, and satellite-based decision making.
GridEyeS is, as such a novel framework for heterogeneous spatiotemporal data fusion. The heart of the GridEyeS framework is a novel machine learning algorithm called the Multi-task Deep Recurrent Neural Network (MDRNN) to discover the relationship between aerial imaginary data, electrical measurement, and meteorology data. Multi-Task learning is a new family of machine learning algorithms in which multiple data models are jointly learned over multiple data sources to improve algorithm performances.
Space Added Value
The space assets utilized in the proposed system will mainly be earth observation data from optical satellites. The main goal is to incorporate satellite images in a streamlined end-to-end system used for drone surveillance planning, resilience assessment and post event restoration planning of electricity grids. Although it is anticipated that Sentinel-1a/1b, 2, and 3 will be the most important imaging sensors in the demonstration, the use of other optical sources will be analysed to demonstrate the complete sensor suite normally used in targeted market. These satellites include SPOT-5 and 6, WorldView 2 and 3 and also data providing from Planet, DigitalGlobe and GeoEye.
Satellite navigation will of also be an important part of the drone operations. GNSS data will be used to combine drone positions at the time of satellite imagery to allowing assimilation of drone and satellite images in the machine learning algorithm. By combining information from different satellite derived or satellite aided observations in a machine learning algorithm, hot-spots of high risk, or areas needing detailed investigations by drone surveillance can be identified.
Satellite data is also indirectly be used by the weather models driving the operational aspects of drone flight planning, pre event resilience assessments, and post event restoration plans.
A series of interviews with responsible for vegetation management have been concluded focusing on incremental value creation compared to existing workflows. Based on the collected information, mock-ups have been completed. The team has acquired a good understanding of the potential market size and entry barriers, as well as the technical feasibility of optimizing vegetation management utilizing and combining high resolution data sources like satellite, UAV, LIDAR, weather and ground observations.