- Python 3
numpymatplotlibfor drawingpandas- only for showing the results in dataframes
To evaluate a layout, see run.py script in src directory, and execute python3 run.py.
See also Toy-Model.ipynb notebook in notebooks directory.
To create a new layout for evaluation and testing using the Tiny Plant, you have to write a list of n=5 heliostat coordinates [xi, yi] in the following format:
[[x1, y1], [x2, y2], ..., [xn, yn]]
and save it as name-of-layout.json to data/layouts directory.
For the layout to be valid, the coordinates have to be in the Tiny Plants field area:
0 <= xi <= 35
0 <= yi <= 10
Heliostats must not be too close, they have to be at least 2 times the radius of a bounding circle appart, this distance is equal to heliostat size:
|(xi, yi) - (xj, yj)| >= 4
And heliostats must not be too close to the receiver
|(xi, yi) - (0, 12)| >= 4
See the existing layouts in data/layouts, for example: grid-layout.json with coordinates for 5 heliostats.
Below are the current results for a few basic layouts evaluated on the Tiny plant using only a few Sun angles and Sun rays in order to run faster:
| Layout | Energy |
|---|---|
| Simulated Annealing | 65.32 |
| Differential Evolution | 65.14 |
| Parabolic | 63.14 |
| SQP | 61.34 |
| Basin-hopping | 61.14 |
| Cobyla | 61.04 |
| Theater | 60.64 |
| Best Random | 59.26 |
| Spiral | 55.62 |
| Grid | 48.22 |
| Tiny | 45.19 |
Increasing the number of angles we consider from 17 to 180, and the number of rays from 5 to 50, we see that some layouts were only exploiting the inaccuracies in the Toy model:
| Layout | Energy |
|---|---|
| Parabolic | 717.42 |
| Differential Evolution | 693.42 |
| Theater | 693.34 |
| Simulated Annealing | 692.04 |
| Cobyla | 685.48 |
| Best Random | 677.48 |
| SQP | 655.44 |
| Basin-hopping | 652.77 |
| Spiral | 593.12 |
| Grid | 534.99 |
| Tiny | 522.43 |