Latest posts by Jose Carballo (see all)
- Smart solar tower power mockup based on deep learning for computer vision - 21 January, 2020
In this article, we will introduce a fully functional smart solar tower mockup. The mockup is based on a new concept to track the Sun and the receiver called Heliot, a smart heliostat approach. This approach is based on computer vision and deep learning for object detection using Tensorflow as machine learning framework.
Have a look at the video below to see the smart solar tower power mockup in action.
- Why Renewable Energy?
- Why Concentrating Solar Power?
- Concentrating Solar Power Technologies
- Solar Tracking System
- Heliot – Smart Heliostat
- Smart Solar Tower Mockup
- Heliostat Mockup: 3D-printed Parts & Hardware
- Solar Tower Mockup: 3D-printed Parts & Hardware
- Deep Learning – Object Detection
- Summary & References
Why Renewable Energy?
Energy makes the world go round. Industry, daily life, transportation, communications, all is powered by energy. Industrial developments and population grow have been exponentially rising our need for energy in order to maintain our energy-dependent lifestyles. Fossil fuels are currently our major power supplies, fossil fuels’ market share ranges from 87% to 76%.
However, rising cost, environmental issues and concerns about sustainability are presently encouraging investment and research in alternative sources of energy that may provide clean, renewable, sustainable and efficient energy.
Moreover, renewables are, by definition, the only way we can keep producing energy for a virtually infinite period of time!
Why Concentrating Solar Power?
Among all renewables, solar energy is the most abundant energy source on Earth. About 93% of solar energy may be theoretically used. The two major solar energy technologies are solar photovoltaic (PV) and solar thermal.
The most efficient way of storing heat in solar thermal energy, than electricity in PV batteries, makes solar thermal energy more appropriate for large-scale energy production facilities due to its dispatchability on demand.
Optical concentration devices are used to achieve maximum solar conversion at little heat loss by concentrating the reflected solar flux onto a reduced-area receiver. This technology is known as Concentrating Solar Thermal (CST)
When CST is applied to electricity production is known as Concentrating Solar Power (CSP) or Solar Thermal Electricity (STE).
Concentrating Solar Power Technologies
Four CSP technologies are presently available:
- Dish / Engine (DE) systems
- Linear Fresnel (LF) systems
- Parabolic Trough (PT) collectors
- Solar Power Tower (SPT) based on heliostats
Solar Power Tower (SPT) is one of the most promising CSP technologies since it yields the highest efficiency and has the greatest scaling-up potential. Solar tower power plants account for 70% of the total capacity of CPS projects under development.
SPT transforms solar energy into thermal energy to later produce electricity. Solar energy concentrators (called heliostats), compounded by mirrors, reflect and concentrate the solar radiation onto a receiver placed at the top of a tower.
The receiver is responsible for absorbing and transforming concentrated radiation into thermal energy. Then the thermal energy is delivered to a heat transfer fluid to be used in a power cycle to produce electricity thanks to a turbine or stored in a tank for later use, for instance at night or under unfavorable meteorological conditions.
Solar Tracking System
Due to the constant movement of the Earth, heliostats must be constantly aligned to reflect and concentrate the incoming solar radiation onto the receiver. This task is carried out by the solar tracking system.
The solar tracking systems keeps the optical axis of the heliostat (A’) aligned with the angle bisector (VA) formed by the solar vector (VS) and the receptor vector (VT). In the figure below, the target is the place where the incoming solar radiation is reflected instead of the receiver. The target is used in STP plants for testing and heliostat calibration purposes.
Traditional solar tracking systems are based on solar position algorithms because the position of the receiver or target is known and the apparent Sun’s trajectory is already well known and given by solar equations.
However, this approach have some drawbacks due to the solar tracking system works without any kind of feedback.
- Harware is expensive because precise equipment must be used to keep heliostats aligned over time.
- The installation cost is high because the traditional approach assumes that heliostats are precisely perpendicular with respect to the ground. The installation is therfore expensive, time consuming and it is difficult to guarantee the required accuracy.
- The maintenance cost is high. There are several factors that can misaligned heliostats: wind, rain, terrain movements, remote earthquake, etc. In practice, each heliostat must be periodically calibrated to cope with misalignments. This task is also time consuming and must be performed during the day over the target, and therefore
This figure shows Ivanpah power plant, the largest STP plant up to date, with 173,500 heliostat, 3 tower and a gross capacity of 393 MW. Can you imagine how expensive it is to calibrate the whole heliostat field in such a large STP plant?
Heliot – Smart Heliostat
Heliot comprises a camera attached to the heliostat surface mirror’s center point that moves with it (although other locations are also possible). The camera provides a plane view of the scene (CP).
The neural network detects the Sun’s (S’) and target’s (T’) (or receiver’s) center points. The middle point (A’) between the Sun and target is the desired heliostat aiming point.
The current heliostat aiming point (A”) is the center point in the plane view. The heliostat is then continuously move to place the current aiming point at the desired aiming point (A” = A’).
Heliot can be used in both CSP and PV technologies.
The main Heliot advatanges are the following.
- Hardware is low cost. The prototype is based on a Raspberry Pi and Raspberry Pi Camera.
- Installation cost reduction. There is no need for an expensive and time-consuming installation procedure since Heliot can compensate misalignments in real time.
- Maintenance cost reduction. No calibration is required since Heliot calibrate itself in real time.
- Forecasting capabilities. The system can detect clouds and estimate their directions and velocities to perform predictions about near future solar radiation and plant performance.
- Detect design issues. The system can detect surrounding heliostat and determine if they produce block or shadow on other heliostats.
The main challenges that Heliot must overcome are:
- Object detection. Heliot must detect the objects of interest in diverse and challenging conditions: lighting conditions, sunset, sunrise, dust, dirt, etc.
- Accuracy. Heliot must reach the desired accuracy to properly aim heliostats, making sure that the differences in distance between the detected objects and the ground truth is under a maximum error.
Smart Solar Tower Mockup
The Smart Solar Tower Mockup implements the Heliot approach previously discussed. This mockup is based on low-cost hardware (Raspberry Pi and Arudino) and is easy to build and assemble because the structure is based on 3D-printed parts.
- Raspberry Pi 3 Model B+
- Raspberry Pi Camera
- Raspberry Pi 3 – 7″ Touchscreen Display
- 2 x Flex Cable for Raspberry Pi (for camera and display)
- 2 x Raspberry Pi 3 B+ Power Supply – 5V 2.5A (for Raspberry Pi and display)
- 3D Printer Filament
- 2 X Servomotor
- LED RGB
- Aurduino UNO microcontroller
- LCD Module Display for microcontrollers
- PV module
- Push Button
- Breadboard Jumper Wires Ribbon Cables
- Concave mirror
These are affiliate links. This means if you click on the link and purchase the promoted item, we will receive a small affiliate commission at no extra cost to you, the price of the product is the same. We would really appreciate your support of our work and website if this is fine for you.
Heliostat mockup: 3D-printed Parts & Hardware
The heliostat design is based on a real heliostat design, where the concave mirror rotates around two axes (horizontal rotation and elevation). The 3D CAD model has been designed in Sketchup. The heliostat design was 3D printed.
The main heliostat hardware comprises two servo motors (horizontal rotation and elevation), a Raspberry Pi 3 Model B+, Raspberry Pi V2 8-megapixel camera, a 7-inch touchscreen display and a status LED.
Solar Tower Mockup: 3D-printed Parts & Hardware
The tower includes a photovoltaic (PV) panel acting as a receiver. A PV panel is used in order to measure the incoming energy.
A microcontroller (Arduino UNO compatible board) reads the PV panel voltage and shows the information in a LCD display. The rotary potentiometer set the LCD brightness and the push button determines which information is shown in the LCD display.
You can find interesting tutorials with open-source example in our blog, if you are interested in learning about Arduino and NodeMCU programming.
Deep Learning – Object Detection
Tensorflow Lite is the machine learning framework. It is open source, lightweight and targets mobile and embedded devices. It enables on-device machine learning inference with low latency and small application size
SSD MobileNet artificial neural network was retrained to detect objects that mimic the Sun (a turn-on LED bulb) and the target or receiver (a small-size PV panel) in STP plants.
We used LabelBox to label the training and validation image sets. Remember that the image set must be as heterogeneous as possible if we want to create a robust model.
The app was developed as described in previous tutorials in order to integrate TensorFlow Lite with Qt/QML for the development of Raspberry Pi apps. We also made used of the Coral USB Accelerator to have a fast response.
These are interesting tutorial with open-source example apps, if you want to learn about how to integrate Tensorflow, Qt and the Coral USB Accelerator on Raspberry pi.
- Raspberry Pi, TensorFlow Lite and Qt/QML: object detection example
- Coral USB Accelerator, TensorFlow Lite C++ API & Raspberry Pi for Edge TPU object detection
Summary & References
We have learned about the Heliot tracking approach and how to build a Smart Solar Tower Mockup. The Smart Solar Tower Mockup comprises a heliostat and a solar tower.
The heliostat is made of 3D-printed parts and low-cost hardware (Raspberry Pi, camera, servo motors, status LED and a touchscreen display).
The solar tower is also made of 3D-printed parts and low-cost hardware (PV panel, rotary potentiometer, push button and an Arduino board).
More information about Heliot and the Solar Tower Pocker mockup is given in our scientific publications in journals and international conferences.
- New approach for solar tracking systems based on computer vision, low-cost hardware, and deep learning – Renewable Energy
- Machine learning for solar trackers – SolarPACES conference
- Solar tower power mockup for the assessment of advanced control techniques – Renewable Energy
I hope you liked this article, please consider to rate it with the stars you can find below, this gives us feedback about our work. If you have any doubt, proposal, comment or issue, just write below, we will try to help 🙂