Metaheuristics — Simulated Annealing

UC + Dispatch · SA

State = (Uᵢ ∈ {0,1}, pᵢ) — both commitment and dispatch are optimised simultaneously. Each move either flips a commitment bit (30%) or perturbs an output level (70%). After every move a feasibility repair projects outputs back to satisfy demand and P-min/P-max bounds. Bad moves are accepted with probability e^(−ΔC/T).

Generator Fleet

1

P-Min

P-Max

a

b

c

Start $

2

P-Min

P-Max

a

b

c

Start $

3

P-Min

P-Max

a

b

c

Start $

4

P-Min

P-Max

a

b

c

Start $

Max cap: 220 MW

C(p) = ap² + bp + c + startup

SA Parameters

Exploration breadth

Stop threshold

0.9 – 0.999

≤ 2000

Convergence — Current vs Best Cost

Run the annealer to watch cost converge.

Best Commitment + Dispatch Found

Best solution will appear here after annealing.