NCS Value-Chain Optimisation
This skill optimises the Norwegian Continental Shelf as one system. It takes the
open-data export graph from neqsim-ncs-infrastructure-network, puts supply on
it for every year, and answers:
- how much each field and discovery should produce each year to maximise value
when all fields share the same trunklines, plants and terminals;
- which elements bind, and what one more MSm3/d or Sm3/d of capacity is worth
there (shadow price);
- how an outage, capacity change, price shift or new tie-in propagates;
- where spare capacity (ullage) opens up over time, and which discoveries use
it best;
- which producing fields could ship more, and whether the field or the network
limits them.
Everything is public screening data. Governed inputs (corporate forecasts, Gassled
tariffs, gas-quality specs, metered flows) enter through a scenario JSON
produced by a private enterprise binding skill (a company writes its own; see
the Enterprise Agent and Skill Repositories guide). This skill never reads company
systems.
When to Use
- "Maximise NCS production/value for 2026–2040 given the pipeline system."
- "What is the most valuable debottleneck on the NCS, and when does it bind?"
- "If Kollsnes runs at 70% in 2027, what is lost and where does gas reroute?"
- "Which discoveries should be tied in first, into which host, and is there
ullage?"
- "Which fields could produce more today, and what limits them?"
- Before a detailed NeqSim host process-model study: to get the rate targets
the host must handle, including tie-ins.
Inputs
- An
NcsNetwork (bundled snapshot or a live rebuild).
- A scenario (
ncs_value_chain_scenario.v1, see scenario.py) with these keys:
years;
prices (gas NOK/Sm3, liquid NOK/Sm3, exit-market factors);
tariffs (default and per arc);
discount_rate;
capacity_overrides;
outages (element, years, available fraction);
forecast_overrides (e.g. corporate forecast per field);
include_discoveries, tieback_max_km, max_paths_per_entity.
All have stated public defaults.
Outputs
NcsValueChainOptimizer(network, scenario).run() returns
ncs_value_chain_result.v1. For each year it holds:
produced per entity (u, gas MSm3/d, liquid Sm3/d);
curtailed, element_flow and utilization;
shadow_price_mnok_per_unit_yr, and exits by market;
unsold (production without a sales route, e.g. reinjection) and
unconnected entities.
It also carries the horizon NPV, the tie-back host of each discovery, the
supply profiles with their method, and the stated assumptions.
opportunities.py turns this into:
bottlenecks: elements ranked by discounted shadow value;
ullage_timeline: spare capacity per element and year;
curtailment: lost production by year and entity;
tiein_ranking: discoveries by accepted discounted revenue, with host and
distance;
stranded_discoveries: no host within the tie-back radius;
production_uplift: extra rate per field, limited by route headroom or
demonstrated capability.
Engineering Method
Supply. Sodir yearly net production up to the base year.
- Forecast: an Arps decline fitted from peak, using
norwegian_continental_shelf_data.fit_arps_decline (built-in exponential
fallback), with a 4 %/yr minimum decline. Fields still on plateau hold their
rate until about 60 % of remaining reserves are produced, then decline at
12 %/yr.
- Cap: cumulative future volume is capped at Sodir remaining reserves.
- Discoveries: sized from Sodir recoverable volumes. Start year = base
year + 2 (decided/approved), + 4 (clarification), + 7 (likely); not
evaluated and unlikely are excluded by default. Profile: 50 % build-up
year, then 12 %/yr of recoverable on plateau for 4 years, then 15 %/yr
decline.
Routing. Up to 20 loop-free paths per entity and medium to markets. A
discovery ties back to the nearest in-service processing host within
tieback_max_km that has an export route.
LP per year.
$$\max \sum_{e,p} \left(P_m f_{exit} - T_p\right) x_{e,p} \quad \text{s.t.} \quad \sum_p x^{m}{e,p} = u_e R^{m}e,; \sum{(e,p) \ni c} x{e,p} \le C_c,; 0 \le u_e \le 1$$
Here $u_e$ couples a field's oil and gas, and $C_c$ is every trunkline and
plant capacity after overrides and outages. Solved with SciPy HiGHS. Duals
of the capacity rows are shadow prices in MNOK/yr per MSm3/d (gas) or per
Sm3/d (liquid).
Horizon value. Annual objectives discounted at discount_rate. Years
are independent: curtailed volume is lost, not deferred (conservative).
Python Usage Pattern
from ncs_infrastructure_network import NcsNetwork
from ncs_value_chain_optimization import (NcsValueChainOptimizer, make_scenario, bottlenecks,
tiein_ranking, production_uplift, ullage_timeline)
from ncs_value_chain_optimization import neqsim_bridge as vb
net = NcsNetwork.load()
sc = make_scenario(years=[2026, 2040],
outages=[{"element": "KOLLSNES", "years": [2027], "available_fraction": 0.7}])
opt = NcsValueChainOptimizer(net, sc)
res = opt.run() # ~1 s for 15 years, 160 entities
bottlenecks(res)[:5] # KOLLSNES binds in 2027 at ~1090 MNOK/yr per MSm3/d
tiein_ranking(res, gas_price=3.0, liquid_price=4500.0)[:10]
production_uplift(opt, res, 2026)
ullage_timeline(res, ["LANGELED_SOUTH__SLEIPNER__EASINGTON", "KOLLSNES"])
# NeqSim hand-offs (pip install neqsim)
vb.value_objective_check(res, sc, 2026) # ValueChainObjective reproduces the LP gross revenue
specs = vb.debottleneck_specs(opt, res, [{"element": "KOLLSNES", "added_capacity": 10, "capex_mnok": 3000}])
vb.run_debottlenecking(specs, sc) # DebottleneckingAdvisor: NPV, B/C, payback
vb.process_model_targets(res, net, "GOLIAT") # rate targets for the optimize-processmodel agent
Validated NeqSim Path
| Step |
NeqSim class |
Bridge |
| Economic basis |
neqsim.process.optimization.valuechain.EconomicParameters |
economic_parameters |
| Revenue cross-check |
ValueChainObjective.evaluate(gasSm3/d, oilSm3/d, powerKw) |
value_objective_check |
| Debottleneck ranking |
DebottleneckingAdvisor + DebottleneckCandidate (year offsets from base year) |
debottleneck_specs, run_debottlenecking |
| Tie-back screening |
fielddevelopment.tieback.TiebackAnalyzer |
ncs_infrastructure_network.neqsim_bridge |
| Trunkline deliverability |
equipment.network.LoopedPipeNetwork |
ncs_infrastructure_network.neqsim_bridge |
| Host process capacity |
ProcessAutomation.findMaxThroughputJson, AgenticProcessOptimizer |
process_model_targets → optimize-processmodel agent |
| Field-life investment timing |
LifeOfFieldOptimizer |
take a host's rate targets as its production basis |
Validation Checklist
Common Mistakes
- Too few paths per entity. With 8 paths the LP curtailed about
19 MSm3/d of Troll gas that the network can carry. Keep ≥ 20.
- Treating DebottleneckCandidate years as calendar years. They are offsets
from the base year; 2027 as a year discounts the benefit to zero.
- Using duals for large capacity steps. A dual is marginal. The bridge
re-solves with the added capacity instead.
- Reading profiles as a forecast. Sodir layer 7300 is history only; the
current year is year-to-date.
- Mistaking zero curtailment for no bottleneck. When supply declines faster
than capacity, nothing binds. Check
ullage_timeline for where tie-ins fit.
Limitations
- Nameplate capacities; no line-pack, pressure-dependent capacity, gas-quality
blending, Gassco nomination rules or NGL product constraints at Kårstø.
- Screening supply forecasts (decline + reserves cap). Replace them with corporate or
operator forecasts through
forecast_overrides for decisions.
- Tariffs and exit-market factors are placeholders until bound from governed
commercial data.
- Years are independent (no deferral, storage or reservoir response to
curtailment).
- Host processing capacity is not open data; the host process model must
confirm tie-in rates.
Related NeqSim Functionality
neqsim-ncs-infrastructure-network: the graph, open-API clients, tie-back and
hydraulics bridges.
neqsim-norwegian-continental-shelf-data: fit_arps_decline, headline facts,
carbon-cost basis.
neqsim-resource-classification-screening: RC maturity behind discovery
phasing.
neqsim-asset-value-npv-screening, neqsim-energy-emissions-screening:
project-level economics and emissions of a chosen tie-in or debottleneck.
- NeqSim
optimize-processmodel agent: host plant confirmation of the rate
targets.
References
- Norwegian Offshore Directorate FactPages / DataService (NLOD 2.0).
- Norwegian Petroleum, "The oil and gas pipeline system" and "Resource accounts".
- Arps, J.J. (1945). Analysis of decline curves. Trans. AIME 160, 228–247.
- Huangfu, Q., Hall, J.A.J. (2018). Parallelizing the dual revised simplex method. Math. Prog. Comp. 10, 119–142 (HiGHS).
1---2name: neqsim-ncs-value-chain-optimization3calculation_basis: "screening"4version: "0.1.0"5description: "Multi-year, capacity-constrained NCS value-chain optimisation: forecasts fields from Sodir history, phases in discoveries, routes oil and gas through real trunklines, plants and terminals with a HiGHS LP, and returns production, curtailment, shadow prices, ullage, tie-in rankings and uplift lists for NeqSim follow-up. USE WHEN: maximising NCS value or production, finding the binding pipeline/plant and its capacity value, testing an outage or tie-in, or fitting discoveries into spare capacity."6last_verified: "2026-09-26"7requires:8 python_packages: ["scipy"]9 java_packages: []10 env: []11 network: []12metadata:13 source: https://skillmd.com/skills/equinor/neqsim-ncs-value-chain-optimization14---1516# NCS Value-Chain Optimisation1718This skill optimises the Norwegian Continental Shelf as one system. It takes the19open-data export graph from `neqsim-ncs-infrastructure-network`, puts supply on20it for every year, and answers:2122- how much each field and discovery should produce each year to maximise value23 when all fields share the same trunklines, plants and terminals;24- which elements bind, and what one more MSm3/d or Sm3/d of capacity is worth25 there (shadow price);26- how an outage, capacity change, price shift or new tie-in propagates;27- where spare capacity (ullage) opens up over time, and which discoveries use28 it best;29- which producing fields could ship more, and whether the field or the network30 limits them.3132Everything is public screening data. Governed inputs (corporate forecasts, Gassled33tariffs, gas-quality specs, metered flows) enter through a scenario JSON34produced by a private enterprise binding skill (a company writes its own; see35the Enterprise Agent and Skill Repositories guide). This skill never reads company36systems.3738## When to Use3940- "Maximise NCS production/value for 2026–2040 given the pipeline system."41- "What is the most valuable debottleneck on the NCS, and when does it bind?"42- "If Kollsnes runs at 70% in 2027, what is lost and where does gas reroute?"43- "Which discoveries should be tied in first, into which host, and is there44 ullage?"45- "Which fields could produce more today, and what limits them?"46- Before a detailed NeqSim host process-model study: to get the rate targets47 the host must handle, including tie-ins.4849## Inputs5051- An `NcsNetwork` (bundled snapshot or a live rebuild).52- A scenario (`ncs_value_chain_scenario.v1`, see `scenario.py`) with these keys:53 - `years`;54 - `prices` (gas NOK/Sm3, liquid NOK/Sm3, exit-market factors);55 - `tariffs` (default and per arc);56 - `discount_rate`;57 - `capacity_overrides`;58 - `outages` (element, years, available fraction);59 - `forecast_overrides` (e.g. corporate forecast per field);60 - `include_discoveries`, `tieback_max_km`, `max_paths_per_entity`.61 All have stated public defaults.6263## Outputs6465`NcsValueChainOptimizer(network, scenario).run()` returns66`ncs_value_chain_result.v1`. For each year it holds:6768- `produced` per entity (`u`, gas MSm3/d, liquid Sm3/d);69- `curtailed`, `element_flow` and `utilization`;70- `shadow_price_mnok_per_unit_yr`, and `exits` by market;71- `unsold` (production without a sales route, e.g. reinjection) and72 `unconnected` entities.7374It also carries the horizon NPV, the tie-back host of each discovery, the75supply profiles with their method, and the stated assumptions.7677`opportunities.py` turns this into:7879- `bottlenecks`: elements ranked by discounted shadow value;80- `ullage_timeline`: spare capacity per element and year;81- `curtailment`: lost production by year and entity;82- `tiein_ranking`: discoveries by accepted discounted revenue, with host and83 distance;84- `stranded_discoveries`: no host within the tie-back radius;85- `production_uplift`: extra rate per field, limited by route headroom or86 demonstrated capability.8788## Engineering Method89901. **Supply.** Sodir yearly net production up to the base year.91 - **Forecast:** an Arps decline fitted from peak, using92 `norwegian_continental_shelf_data.fit_arps_decline` (built-in exponential93 fallback), with a 4 %/yr minimum decline. Fields still on plateau hold their94 rate until about 60 % of remaining reserves are produced, then decline at95 12 %/yr.96 - **Cap:** cumulative future volume is capped at Sodir remaining reserves.97 - **Discoveries:** sized from Sodir recoverable volumes. Start year = base98 year + 2 (decided/approved), + 4 (clarification), + 7 (likely); not99 evaluated and unlikely are excluded by default. Profile: 50 % build-up100 year, then 12 %/yr of recoverable on plateau for 4 years, then 15 %/yr101 decline.1022. **Routing.** Up to 20 loop-free paths per entity and medium to markets. A103 discovery ties back to the nearest in-service processing host within104 `tieback_max_km` that has an export route.1053. **LP per year.**106107 $$\max \sum_{e,p} \left(P_m f_{exit} - T_p\right) x_{e,p} \quad \text{s.t.} \quad \sum_p x^{m}_{e,p} = u_e R^{m}_e,\; \sum_{(e,p) \ni c} x_{e,p} \le C_c,\; 0 \le u_e \le 1$$108109 Here $u_e$ couples a field's oil and gas, and $C_c$ is every trunkline and110 plant capacity after overrides and outages. Solved with SciPy HiGHS. Duals111 of the capacity rows are shadow prices in MNOK/yr per MSm3/d (gas) or per112 Sm3/d (liquid).1134. **Horizon value.** Annual objectives discounted at `discount_rate`. Years114 are independent: curtailed volume is lost, not deferred (conservative).115116## Python Usage Pattern117118```python119from ncs_infrastructure_network import NcsNetwork120from ncs_value_chain_optimization import (NcsValueChainOptimizer, make_scenario, bottlenecks,121 tiein_ranking, production_uplift, ullage_timeline)122from ncs_value_chain_optimization import neqsim_bridge as vb123124net = NcsNetwork.load()125sc = make_scenario(years=[2026, 2040],126 outages=[{"element": "KOLLSNES", "years": [2027], "available_fraction": 0.7}])127opt = NcsValueChainOptimizer(net, sc)128res = opt.run() # ~1 s for 15 years, 160 entities129130bottlenecks(res)[:5] # KOLLSNES binds in 2027 at ~1090 MNOK/yr per MSm3/d131tiein_ranking(res, gas_price=3.0, liquid_price=4500.0)[:10]132production_uplift(opt, res, 2026)133ullage_timeline(res, ["LANGELED_SOUTH__SLEIPNER__EASINGTON", "KOLLSNES"])134135# NeqSim hand-offs (pip install neqsim)136vb.value_objective_check(res, sc, 2026) # ValueChainObjective reproduces the LP gross revenue137specs = vb.debottleneck_specs(opt, res, [{"element": "KOLLSNES", "added_capacity": 10, "capex_mnok": 3000}])138vb.run_debottlenecking(specs, sc) # DebottleneckingAdvisor: NPV, B/C, payback139vb.process_model_targets(res, net, "GOLIAT") # rate targets for the optimize-processmodel agent140```141142## Validated NeqSim Path143144| Step | NeqSim class | Bridge |145|------|--------------|--------|146| Economic basis | `neqsim.process.optimization.valuechain.EconomicParameters` | `economic_parameters` |147| Revenue cross-check | `ValueChainObjective.evaluate(gasSm3/d, oilSm3/d, powerKw)` | `value_objective_check` |148| Debottleneck ranking | `DebottleneckingAdvisor` + `DebottleneckCandidate` (year offsets from base year) | `debottleneck_specs`, `run_debottlenecking` |149| Tie-back screening | `fielddevelopment.tieback.TiebackAnalyzer` | `ncs_infrastructure_network.neqsim_bridge` |150| Trunkline deliverability | `equipment.network.LoopedPipeNetwork` | `ncs_infrastructure_network.neqsim_bridge` |151| Host process capacity | `ProcessAutomation.findMaxThroughputJson`, `AgenticProcessOptimizer` | `process_model_targets` → `optimize-processmodel` agent |152| Field-life investment timing | `LifeOfFieldOptimizer` | take a host's rate targets as its production basis |153154## Validation Checklist155156- [ ] Every year `status == "optimal"`; first-year shipped gas within a few % of157 the last historical year (the 2026 base case ships ~320 MSm3/d).158- [ ] `max(utilization) <= 1` in every year.159- [ ] `value_objective_check` relative difference below 1e-6 (the NeqSim160 economics and the LP agree).161- [ ] Shadow prices ≈ netback × 365 on binding gas elements (sanity: (3.0 − 0.12)162 × 365 ≈ 1051 MNOK/yr per MSm3/d).163- [ ] Discoveries used in rankings carry status and RC; start years reviewed.164- [ ] Prices and tariffs are labelled public defaults unless bound from165 governed data.166167## Common Mistakes168169- **Too few paths per entity.** With 8 paths the LP curtailed about170 19 MSm3/d of Troll gas that the network can carry. Keep ≥ 20.171- **Treating DebottleneckCandidate years as calendar years.** They are offsets172 from the base year; 2027 as a year discounts the benefit to zero.173- **Using duals for large capacity steps.** A dual is marginal. The bridge174 re-solves with the added capacity instead.175- **Reading profiles as a forecast.** Sodir layer 7300 is history only; the176 current year is year-to-date.177- **Mistaking zero curtailment for no bottleneck.** When supply declines faster178 than capacity, nothing binds. Check `ullage_timeline` for where tie-ins fit.179180## Limitations181182- Nameplate capacities; no line-pack, pressure-dependent capacity, gas-quality183 blending, Gassco nomination rules or NGL product constraints at Kårstø.184- Screening supply forecasts (decline + reserves cap). Replace them with corporate or185 operator forecasts through `forecast_overrides` for decisions.186- Tariffs and exit-market factors are placeholders until bound from governed187 commercial data.188- Years are independent (no deferral, storage or reservoir response to189 curtailment).190- Host processing capacity is not open data; the host process model must191 confirm tie-in rates.192193## Related NeqSim Functionality194195- `neqsim-ncs-infrastructure-network`: the graph, open-API clients, tie-back and196 hydraulics bridges.197- `neqsim-norwegian-continental-shelf-data`: `fit_arps_decline`, headline facts,198 carbon-cost basis.199- `neqsim-resource-classification-screening`: RC maturity behind discovery200 phasing.201- `neqsim-asset-value-npv-screening`, `neqsim-energy-emissions-screening`:202 project-level economics and emissions of a chosen tie-in or debottleneck.203- NeqSim `optimize-processmodel` agent: host plant confirmation of the rate204 targets.205206## References207208- Norwegian Offshore Directorate FactPages / DataService (NLOD 2.0).209- Norwegian Petroleum, "The oil and gas pipeline system" and "Resource accounts".210- Arps, J.J. (1945). Analysis of decline curves. Trans. AIME 160, 228–247.211- Huangfu, Q., Hall, J.A.J. (2018). Parallelizing the dual revised simplex method. Math. Prog. Comp. 10, 119–142 (HiGHS).