Add production-ready DART market module

.github/workflows/TestOnPRs.yml

@@ -18,12 +18,18 @@ concurrency:
 
 jobs:
   test:
+    strategy:
+      fail-fast: false
+      matrix:
+        version:
+          - "1.9"
+          - "1"
     uses: ./.github/workflows/ReusableTest.yml
     with:
       os: ubuntu-latest
-      version: "1"
+      version: ${{ matrix.version }}
       arch: x64
       allow_failure: false
-      run_codecov: true
+      run_codecov: ${{ matrix.version == '1' }}
     secrets:
       codecov_token: ${{ secrets.CODECOV_TOKEN }}

CHANGELOG.md

@@ -7,6 +7,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- Lightweight in-memory DART SCUC, SCED, generator N-1, rolling, and settlement
+  functionality.
+
 ### Changed
 
 - Add `clean_energy_policy = 2` for local annual RPS compliance without

Project.toml

@@ -14,7 +14,6 @@ DataStructures = "864edb3b-99cc-5e75-8d2d-829cb0a9cfe8"
 Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
 Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"
 GLPK = "60bf3e95-4087-53dc-ae20-288a0d20c6a6"
-Gurobi = "2e9cd046-0924-5485-92f1-d5272153d98b"
 HiGHS = "87dc4568-4c63-4d18-b0c0-bb2238e4078b"
 JuMP = "4076af6c-e467-56ae-b986-b466b2749572"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
@@ -24,6 +23,7 @@ YAML = "ddb6d928-2868-570f-bddf-ab3f9cf99eb6"
 
 [weakdeps]
 CPLEX = "a076750e-1247-5638-91d2-ce28b192dca0"
+Gurobi = "2e9cd046-0924-5485-92f1-d5272153d98b"
 SCIP = "82193955-e24f-5292-bf16-6f2c5261a85f"
 
 [extensions]

README.md

@@ -19,7 +19,7 @@ The **Holistic Optimization Program for Electricity (HOPE)** model is a transpar
 
 1. `GTEP` mode: a generation & transmission expansion planning model
 2. `PCM` mode: a production cost model
-3. `DART` mode: a SCUC/SCED market model for simulating day-ahead and real-time markets
+3. `DART` module: an individual-generator SCUC/SCED model for day-ahead and real-time markets
 4. `OPF` mode: (under development): an optimal power flow model
 5. `HOPE-AI` mode: an AI agent helps connect all HOPE modules and enables complex modeling workflows. The current HOPE-AI framework is powered by [PowerAgent](https://github.com/Power-Agent), while more specialized agents are under development.
 

docs/make.jl

@@ -34,6 +34,7 @@ pages = OrderedDict(
         "Notation" => "notation.md",
         "GTEP" => "GTEP.md",
         "PCM" => "PCM.md",
+        "DART" => "DART.md",
     ],
     "Input Data Explanation" =>
         ["GTEP Inputs" => "GTEP_inputs.md", "PCM Inputs" => "PCM_inputs.md"],

docs/src/95-reference.md

@@ -18,3 +18,9 @@ Pages = ["95-reference.md"]
 ```@autodocs
 Modules = [HOPE]
 ```
+
+## DART module
+
+```@autodocs
+Modules = [HOPE.DART]
+```

docs/src/DART.md

@@ -0,0 +1,106 @@
+```@meta
+CurrentModule = HOPE.DART
+```
+
+# DART
+
+DART is HOPE's lightweight day-ahead and real-time operations module. It is an
+in-memory API: callers construct typed Julia inputs and receive typed Julia
+results. It does not introduce another case-file schema.
+
+The V1 model contains:
+
+- hourly individual-generator day-ahead SCUC;
+- rolling real-time SCED, normally at five-minute resolution;
+- lossless nodal PTDF transmission constraints;
+- regulation, spinning, and quick-start non-spinning reserves;
+- strict generator N-1 corrective redispatch by default;
+- chronological storage with mutually exclusive charge/discharge operation; and
+- day-ahead/real-time energy, reserve, and uplift settlements.
+
+The implementation is contained in `src/DART.jl`. The main entry points are
+`solve_dart_scuc`, `solve_dart_sced`, `run_dart_rolling`, and
+`calculate_dart_settlements`.
+
+## Formulation mapping
+
+| Formulation group | Implementation |
+|:--|:--|
+| SCUC-OBJ, UC-1--UC-4 | Commitment, startup/shutdown, minimum up/down time, and commitment costs |
+| GEN-1--GEN-7, RAMP-1--RAMP-2 | Generator bounds, transition limits, hourly ramps, and reserve headroom |
+| REN-1--REN-2 | Time-varying availability; zero-cost curtailment is eliminated algebraically |
+| RES-1--RES-5 | Product capability, response time, requirements, and quick-start eligibility |
+| STO-1--STO-6 | Exclusive charge/discharge modes, SOC chronology, and reserve-energy deliverability |
+| NET-1--NET-4 | Nodal injection, balance, PTDF flow, and line limits |
+| SEC-1--SEC-5 | One corrective scenario per eligible generator |
+| RT-1--RT-3 | Fixed DA commitment and energy-only real-time ramping |
+| COUP-1 | Binding RT dispatch and SOC are carried into the next solve |
+
+V1 deliberately uses system-wide reserve requirements, nodal load shedding,
+generator contingencies, and the lossless PTDF network option. Line outages,
+zonal transport, demand response, and policy constraints remain outside this
+small operational core.
+
+Contingency load shedding is disabled by default. Set
+`allow_emergency_contingency_shed = true` in `DARTConfig` only when a penalized
+soft-security solve is preferable to infeasibility.
+
+Quick-start reserve is available only while a unit is offline, in service,
+within its response-time qualification, beyond any remaining minimum-down
+obligation, and within its availability- and forced-outage-derated capacity.
+
+Storage charge and discharge are separated by a binary operating-mode variable
+in both SCUC and SCED. A horizon-end SOC can be imposed by setting
+`terminal_storage_soc_mwh` in `DARTForecast`. It is optional because forcing a
+terminal target at the end of every short rolling RT look-ahead can distort the
+binding dispatch. During rolling simulation, a supplied target applies only to
+a slice that reaches the end of the corresponding input forecast.
+
+## Operational safeguards
+
+`DARTDispatchResult` records the solver termination status, primal status,
+solve time, and relative optimality gap. An optimal result is always accepted.
+By default, a time-limited award is also accepted when it has a feasible primal
+solution and a relative gap no greater than one percent. Configure this with
+`accept_feasible_time_limit` and `maximum_relative_gap`. The fixed-integer LP
+used to produce settlement prices must still solve to optimality.
+
+Explicit generator-contingency variables scale with the number of generators,
+intervals, and monitored outages. DART estimates that block before creating it
+and rejects a solve above `maximum_security_variables` (two million by
+default). Reduce the look-ahead or the set of generators marked
+`contingency_eligible`, or deliberately raise the limit after confirming the
+available memory and solver capacity.
+
+Inputs are checked before model construction for dimensions, finite values,
+physical limits, known reserve products, valid state values, and configuration
+consistency.
+
+## Settlements
+
+Energy settles at fixed-integer nodal prices. Because contingency deliverability
+can make reserve value resource-specific, V1 pays reserve awards at each
+resource's submitted reserve offer instead of reporting a misleading uniform
+security-reserve price. The result field
+`reserve_requirement_shadow_price_per_mw_hour` is the informational dual of the
+system reserve-requirement constraint; settlements do not use it. The
+settlement output identifies this rule as `:pay_as_bid`.
+
+Settlement input must contain one binding column per result, a valid
+chronological RT-to-DA mapping, and exactly one DA interval of RT duration for
+each DA award. Reserve and uplift charges are allocated in proportion to
+real-time served energy. If there is no served energy, the amounts are reported
+explicitly as `unallocated_reserve_charge` and
+`unallocated_uplift_charge`; they are included in the settlement balance rather
+than silently disappearing.
+
+## Minimal example
+
+The ModelCases repository contains `DART_two_bus_example.jl`, a programmatic
+example with no generated or committed data files.
+
+From the HOPE repository root, run it with:
+
+```bash
+julia --project=. ModelCases/DART_two_bus_example.jl
+```

docs/src/model_introduction.md

@@ -60,3 +60,9 @@ Both `GTEP` and `PCM` support:
   - `2`: REG + SPIN + NSPIN
 - `summary_table`:
   - `1`: write summary analytics to `output/Analysis/Summary_*.csv`
+
+## DART Module
+
+The lightweight in-memory DART API provides individual-generator day-ahead
+SCUC, rolling real-time SCED, generator N-1 security, and two-settlement
+accounting. See [DART](@ref).