Load Flow

If bus currents were known, load flow would be one matrix inversion. But a load draws so many megawatts, not so many amps — and S = VI* puts the unknown voltage on both sides of the equation. There is no closed form, so iteration is a necessity rather than a convenience.

The slack bus exists because transmission losses cannot be known until the flows are solved, so one generator must be left free to absorb the difference. It also fixes the angle reference, since only angle *differences* are physically meaningful.

Sparse matrix techniques are what made Newton-Raphson affordable for large networks. Without them the Jacobian solve would dominate everything.

Because X ≫ R on transmission lines, real power depends chiefly on the angle across a line and reactive power on the voltage magnitude. Dropping the cross terms and freezing the Jacobian gives the fast decoupled method — a few more iterations, each far cheaper, fast enough for real-time operation.

The same decoupling is why frequency control and voltage control are separate systems in a control room, operating on different timescales with different equipment.

• Bus voltages within limits, typically 0.95 to 1.05 p.u.
• Line flows within thermal and stability ratings.
• Generators within their P and Q capability — and Q limits bind more often than P limits.

A PV bus that hits its reactive limit is reclassified as PQ mid-solution and its voltage allowed to fall. That is a real operational event — the generator has run out of excitation — not a numerical convenience.

Non-convergence is not always numerical. Past the nose of the P–V curve there is no solution at all, and the physical system undergoes voltage collapse.

Several major blackouts were preceded by studies that would not converge, which makes the failure itself a warning rather than an inconvenience. The whole study is repeated with each single element removed — the N−1 contingency requirement — precisely to find these cases in advance.