Economic Dispatch is the relentless mathematical optimization engine that minimizes the cost of running a power grid. Given 500 different power plants and a constantly shifting city load, a supercomputer runs complex linear programming algorithms every 5 minutes to determine exactly which plants to turn up and which to turn down, ensuring the absolute cheapest megawatts are generated without overloading any physical transmission wires.
Every power plant has a different cost curve. A solar farm's marginal cost is $0/MWh. A combined-cycle gas plant is $40/MWh. A decrepit oil peaker is $200/MWh. Economic Dispatch algorithmically stacks these plants from cheapest to most expensive (the Merit Order). The algorithm simply moves up the stack until supply perfectly equals demand. However, the modern grid runs Security Constrained Economic Dispatch (SCED). The supercomputer must factor in the physical physics of the wires. It might want to dispatch the cheap $0/MWh wind farm in the north, but if the mathematical power flow (Load Flow Analysis) shows that doing so would melt a 220kV transmission line, the SCED algorithm is forced to ignore the cheap wind and dispatch an expensive $200/MWh gas plant near the city to respect the physical wire limits.
Bid Ingestion
Generators submit their massive multi-part cost curves (Start-up cost, No-load cost, Marginal energy cost) to the grid operator's market engine.
State Estimation
The EMS takes a real-time snapshot of the grid's topology, reading thousands of circuit breakers to know exactly which wires are connected and healthy.
SCED Execution
The optimization solver (often using IBM CPLEX or Gurobi) crunches the math in under 3 minutes, spitting out exact MW setpoints (Base Points) for every plant, minimizing total cost while ensuring zero thermal limits are breached.
- Computational Limits: Running a SCED with 10,000 nodes and 50,000 transmission constraints every 5 minutes pushes the limits of modern supercomputing. Integrating millions of decentralized home batteries (DERs) threatens to break the algorithm entirely.
- Non-Convex Cost Curves: Gas turbines are not perfectly linear; they have weird efficiency 'valleys' and 'valves', forcing the algorithm into incredibly complex, slow-to-solve mathematical approximations.