-
Notifications
You must be signed in to change notification settings - Fork 0
/
particle_swarm.py
205 lines (183 loc) · 8.18 KB
/
particle_swarm.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
import numpy as np
import matplotlib.pyplot as plt
from objective_function import ObjectiveFunction
from typing import Union
class Particle:
"""PSO algorithm Particle class"""
def __init__(
self,
objective_function: ObjectiveFunction,
position: np.ndarray,
bounds: Union[np.ndarray, None],
load: Union[np.ndarray, None],
) -> None:
self.dim = bounds.shape[1]
self.bounds = bounds
self.load = load
self.objective_function = objective_function
self.position = position
self.adjust_to_constraints()
self.fitness = self.evaluate_fitness()
self.velocity = np.random.uniform(-1, 1, self.dim)
self.pbest = np.copy(self.position)
def evaluate_fitness(self):
"""Calculates the fitness of a a particle depending on its posistion and an objective function"""
return self.objective_function.evaluate(self.position)
def update_velocity(self, w, c1, c2, gbest):
"""Updates the velocity of the current particle."""
r1 = np.random.random()
r2 = np.random.random()
cogn_velocity = c1 * r1 * (self.pbest - self.position)
social_velocity = c2 * r2 * (gbest - self.position)
self.velocity = w * self.velocity + cogn_velocity + social_velocity
def adjust_to_constraints(self):
if self.bounds is not None:
self.position = np.clip(self.position, self.bounds[0], self.bounds[1])
if self.load is not None:
self.position[:-3] = self.load - (
self.position[-3] + self.position[-2] + self.position[-1]
)
self.position[self.position < 0] = 0
def update_position(self):
"""Updates the positions of the current particle while respecting load and bounds constraint."""
self.position = self.position + self.velocity
self.adjust_to_constraints()
class PSO:
"""PSO Algorithm Class Definition"""
def __init__(
self,
n_particles: int,
max_iters: int,
objective_function: ObjectiveFunction,
bounds: Union[np.ndarray, None],
load: Union[np.ndarray, None],
minimize: bool = True,
) -> None:
self.minimize = minimize
self.n_particles = n_particles
self.bounds = bounds
self.load = load
self.max_iters = max_iters
self.objective_function = objective_function
self.particles = self._init_particles()
self.gbest = min(self.particles, key=lambda particle: particle.fitness).pbest
self.run_results = []
self.history = []
def _init_particles(self):
"""Initializes the positions of the particles in the PSO algorithm by sampling from a uniform distribution"""
particles = []
for _ in range(self.n_particles):
position = np.random.uniform(self.bounds[0], self.bounds[1])
particle = Particle(
self.objective_function, position, self.bounds, self.load
)
particles.append(particle)
return particles
def _step(self, w, c1, c2):
"""Computes ones step of the SPO algorithm for n particles: evalute -> check -> update"""
for idx in range(self.n_particles):
self.particles[idx].update_velocity(w, c1, c2, self.gbest)
self.particles[idx].update_position()
new_fitness = self.particles[idx].evaluate_fitness()
if self.minimize is False:
if new_fitness > self.particles[idx].fitness:
self.particles[idx].fitness = new_fitness
self.particles[idx].pbest = np.copy(self.particles[idx].position)
if new_fitness > self.objective_function.evaluate(self.gbest):
self.gbest = np.copy(self.particles[idx].position)
else:
if new_fitness < self.particles[idx].fitness:
self.particles[idx].fitness = new_fitness
self.particles[idx].pbest = np.copy(self.particles[idx].position)
if new_fitness < self.objective_function.evaluate(self.gbest):
self.gbest = np.copy(self.particles[idx].position)
self.history.append(
{
"gbest": self.gbest,
"particles": self.particles,
"fitness": self.objective_function.evaluate(self.gbest),
}
)
def run(self, w, c1, c2):
self.run_results = []
self.history = []
"""Executes a full run of the PSO algorithm on all particles for a specified number of iterations and prints resutls."""
print("Executing PSO algorithm")
print(f"Number of particles: \t{self.n_particles}")
print(f"Number of Iterations: \t{self.max_iters}")
print(f"w:\t{w}")
print(f"c1:\t{c1}")
print(f"c2:\t{c2}")
print("\n")
for idx in range(self.max_iters):
self._step(w, c1, c2)
print(
f"\tStep {idx}/{self.max_iters}:\t gbest: [{self.gbest[0]:.2f}...{self.gbest[30]:.2f}...{self.gbest[-1]:.2f}] fitness: {self.objective_function.evaluate(self.gbest):.2f}"
)
self.run_results.append(self.objective_function.evaluate(self.gbest))
print("\nPSO run finished.")
def results(self):
"""Prints the results of the PSO algorithm after a successful run."""
print(f"PSO Run Results:\n")
print(
f"\tOptimal PV size: {self.gbest[-2]:.2f}\t\t Interval: [{self.bounds[0][-2]:.2f},{self.bounds[1][-2]:.2f}]"
)
print(
f"\tOptimal ESS Capacity: {self.gbest[-1]:.2f}\t Interval: [{self.bounds[0][-1]:.2f},{self.bounds[1][-1]:.2f}]"
)
print(
f"\tOptimal SF Capacity: {self.gbest[-3]:.2f}\t Interval: [{self.bounds[0][-3]:.2f},{self.bounds[1][-3]:.2f}]"
)
print(
f"\tOptimal values for PD: \t\tInterval: [{self.bounds[0][0]:.2f},{self.bounds[1][0]:.2f}]\n"
)
for idx, value in enumerate(self.gbest[:-3]):
loads = f"\t\tLoad: {self.load[idx]:.4f}" if self.load is not None else ""
msg = f"\t\t PD_{idx+1}: {value:.4f}" + loads
print(msg)
def plot_run(self):
x = range(self.max_iters)
y = self.run_results
fig = plt.figure()
plt.plot(x, y)
plt.xlabel("Step Number")
plt.ylabel("Fitness of gbest")
plt.title("PSO Run plot.")
plt.show()
def plot_particle_movement(self, particle_idx=Union[str, int]):
"""Plots the movement of a chosen particle during the run in a 3D plane"""
# pour ajouter Parreto, tu changes les composants du vecteur position ([pd_1, pd_2, ..., sf, pv, ess]: 3843)
# vect = [pd_1, .., emission, cost]
# x = vect[-1] cost
# y = vect[-2] emission
if particle_idx == "gbest":
history = [hist_dict["gbest"] for hist_dict in self.history]
else:
history = [hist_dict["particles"][particle_idx].position for hist_dict in self.history]
fig = plt.figure()
ax = plt.axes(projection="3d")
for idx, position in enumerate(history):
x = position[-2]
y = position[-1]
z = self.objective_function.evaluate(position)
# pour changer le type du graphe
ax.scatter(
x,
y,
z,
)
ax.text(x, y, z, f"{idx+1}", color="black", fontsize=8)
ax.set_xlabel("PV")
ax.set_ylabel("ESS")
ax.set_zlabel("Fitness")
def __str__(self):
"""Desciption of the PSO algorithm object."""
msg = f"PSO Algorithm:\n"
msg += f"\tNumber of particles: {self.n_particles}\n"
msg += f"\tMaximum Number of Iterations: {self.max_iters}\n"
msg += f"\tsolution Space dimension: {self.bounds.shape[1]}\n"
msg += "\tObjective: Minimize\n" if self.minimize else "Objective: Maximize\n"
msg += "\n\tParticles: \n"
for idx, part in enumerate(self.particles):
msg += f"\t\t|Particle {idx+1}:\tInitial Fitness: {part.fitness}\t Initial Position: [{part.position[0]:.4f} ... {part.position[-1]:.4f}]\n"
return msg