surge-dispatch

Unified dispatch and commitment workflows for Surge.

This crate exposes the typed dispatch request/result surface used for SCED, time-coupled dispatch, commitment optimization, and related secure dispatch workflows across the workspace.

Public API

DispatchModel::prepare plus solve_dispatch is the canonical workflow. Requests are described in terms of three explicit study axes:

• formulation: DC or AC
• coupling: period-by-period or time-coupled
• commitment policy: all committed, fixed, optimize, or additional

Results are keyed-first. The main output is DispatchSolution, which carries:

• resources and buses catalogs with stable ids
• per-period resource_results, bus_results, reserve_results, and constraint_results
• keyed horizon summaries in resource_summaries
• exact objective_terms ledgers plus a persisted audit block that must reconcile to total_cost

use surge_dispatch::{
    DispatchModel, DispatchRequest, DispatchTimeline, solve_dispatch,
};

let model = DispatchModel::prepare(&network)?;
let request = DispatchRequest::builder()
    .dc()
    .period_by_period()
    .all_committed()
    .timeline(DispatchTimeline::hourly(1))
    .build();

model.validate_request(&request)?;

let solution = solve_dispatch(&model, &request)?;
let period = &solution.periods()[0];
let top_bus_lmp = period.bus_results()[0].lmp;
let first_resource = &period.resource_results()[0];
# Ok::<(), surge_dispatch::DispatchError>(())

The crate still contains raw solver-shaped internals for SCED and SCUC, but those are implementation details rather than the product API.

Audit Contract

DispatchSolution is now ledger-first. Every persisted dollar amount is either:

• an exact objective_terms contribution
• a derived rollup of those exact terms

The serialized audit block records whether the solution passed exact reconciliation, whether any residual terms remain, and the concrete mismatch rows when it did not. Export paths should treat audit.audit_passed = false as a failed final artifact rather than a soft warning.

Attribution

The variables, constraints, and objective structure implemented here follow the problem formulation published for the ARPA-E Grid Optimization (GO) Competition Challenge 3. See NOTICE and the References page for the full citations (PNNL-35792 / arXiv:2411.12033).