15.h Improving the NIO Heliostat Field Optical Error Characterization Method and Tools

The efficiency of a utility scale Concentrated Solar Power using the central tower concept with thousands of heliostats in a surround field depends heavily on the optical characteristics of the mirrors and of their tracking accuracy. These characteristics are subject to degradation due to poor construction and installation practices, to environmental factors such as mechanical wear and hardware deterioration, to the limitation of pointing calibration techniques and some other factors. It is therefore critical to be able to assess the optical quality of the mirrors and provide full knowledge to plant managers about which heliostats are underperforming and should be corrected appropriately. The various methods that have been developed to date do not permit a rapid characterization of the heliostats optical quality and consequently negatively affect the solar field performance and cause substantial losses of revenue. The Non-Intrusive Optical (NIO) method is addressing this specific issue by providing an approach that enables a fast heliostats optical characterization without interfering with the plant operation. The proposed solution of the stated problem is to use Unmanned Aerial Vehicle (UAV) to conduct an efficient and accurate optical survey of large-scale heliostat field. The imagery (videos and pictures) collected from the camera carried by the drone, capture the reflection of the central tower on the mirror facets of the heliostats. Using advanced computer vision algorithms make possible the computation of the optical performance of the heliostat through the analysis of the distortion observed on the pictures. In the proposed Phase I, we intend to initiate the maturation of the NIO toolset by accelerating both the data collection and the data processing in order to achieve the true promise of this technology. Our work will focus on multiple areas of improvements for the NIO approach: first we will leverage the power of General Purpose computing on Graphical Processing Unit (GPGPU), which relies on the huge parallelism of current generation of graphic cards that will allow the NIO algorithms to be 30 to 100 faster than the current demonstration implementation; second, we will integrate machine learning techniques for part of the complex computational workflow to enable the automated identification of key images features. Lastly, we intend to assess the possibility to use swarms of drones for accelerating the solar field survey operations. Phase I seeks to demonstrate the feasibility of the proposed improvements, while Phase II would bring the NIO toolset to full commercial phase. With the current ambitious US and overall world push to a carbon neutral power generation, the central tower solar power plants technology is promised a bright future. With its ability to store energy economically in molten salt tanks, and the capability to reach high temperature for multiple purposes (more efficient power generation, Solar Thermochemical, Hydrogen production, thermal output) the technology can play an important part in the solar future. To reach its potential a key parameter is the need to characterize the tens of thousands mirrors’ optical performance and provide plant operators with the knowledge to correct poor heliostat performers. Today about two dozen central tower power plants are in operation in the world, with large number of new plants being currently in the design phase. The improved NIO toolset that we intend to bring to its commercial phase will be able to provide critical services to these solar plants. The use of this technology can help improve the solar plant efficiency by a few percentage points, thus lowering the LCOE.