bhavya1705/windfarmoptimization

Windfarm Optimization using Particle Swarm Optimization (Done using PySwarms) and Circle Packing.

WindFarmOptimization

Using Particle Swarm Optimization and Circle Packing

Approach 1:

Wind Farm Optimization Using MINIMIZATION TECHNIQUES.

Problem-

To optimize placement of Windmills in a Wind Farm constrained by perimeter, distance and subject to Wake Effect.

Solution-

To solve the given problem, the given sample location file was used as starting value. The library py-wake was used to generate wake field. The given Farm_evaluator_vecc file was reconstructed for taking input an array of dimension 100, and used as cost function for a minimization problem. The return value of Farm_evaluator_vecc was inverted.

Methods used for Minimization-

• Particle Swarm Optimization using PySwarm[1]
• Scipy Optimizer- “nelder-mead”[2]
• Genetic Algorithm Using PyGAD and GeneticAlgorithm Libraries[3]
• Simulated Annealing Using Mlrose library[4]
• Random Hill Climb/Hill Climb Using Mlrose library[5]

In initial stages of Optimization Particle Swarm was used to optimize the cost function with initail value set to Turbine_loc_test as initial value. With over ~50,000 iterations, the Average Energy Production improved from ~505 GWh to ~515 GWh. At this stage net increase in gain from Particle Swarm decreased, possible stagnation in local minima. To further optimize the cost function, A combination of Scipy Optimizer and Particle Swarm was used, with each taking the output of the other as initial value. This followed for over ~10,000 iterations with increase in Energy to ~520GWh. With no more improvement from Particle Swarm, Scipy Optimizer was singularly run with over ~30,000 iterations, which increased Average Power from ~520GWh, to ~525GWh. No further progress was made using Scipy Optimizer. An increase of ~1GWh was done using Hill Climb method in Mlrose library in python.

Fig.1. Change in Energy V/S CSV Files Created

Fig.2. Visualization using Py-Wake

Fig.3. Placement of Turbines

Other Techniques used:

• Wind farm layout optimization problem (WFLOP) SUGGA Python toolbox
  (https://github.com/JuXinglong/WFLOP_SUGGA_Python) -Allows for broad level Optimization, does not do changes locally.

References

1. https://docs.scipy.org/doc/scipy/reference/optimize.minimize-neldermead.html#optimize-minimize-neldermead
2. https://mlrose.readthedocs.io/en/stable/source/tutorial1.html
3. https://pyswarms.readthedocs.io/en/latest/
4. https://pygad.readthedocs.io/en/latest/
5. https://pypi.org/project/geneticalgorithm/
6. https://topfarm.pages.windenergy.dtu.dk/PyWake/notebooks/Quickstart.html
7. https://github.com/JuXinglong/WFLOP_SUGGA_Python

Material that require Looking into

1. https://github.com/byuflowlab/PlantEnergy
2. https://github.com/sohailrreddy/WindFLO