Sub-Module 5.5-A
TIMES-NZ and the NZ Energy Modelling Ecosystem
What TIMES-NZ Is and How It Works
TIMES-NZ is a national energy system model built on the IEA ETSAP TIMES framework, developed and maintained by EECA in partnership with the BusinessNZ Energy Council (BEC), with the model files and documentation openly published on GitHub and ReadTheDocs. It finds the least-cost configuration of the New Zealand energy system over a declared planning horizon, subject to technology, resource, demand, and policy constraints. It covers all energy sectors simultaneously — electricity generation, industrial process heat, space heating, transport, and agriculture — making it the most comprehensive model of the national energy system available in New Zealand and the primary tool informing EECA, MfE, and the Climate Change Commission’s long-run pathway analysis. EECA describes its role precisely: the model ‘doesn’t tell us how to run the energy market. Instead, it can help inform conversations about important trade-offs and uncertainties.’ This framing positions TIMES-NZ as a national conversation-framing tool rather than an operational decision system — exactly the analytical gap that the site-level layer of the DCM framework is designed to fill. The model also reflects ‘the impacts of factors largely beyond New Zealand’s control — for example the cost of energy-related imports,’ which identifies an explicit class of exogenous uncertainty that remains unresolved at the national level and must be represented in the FutureArtefact ensemble at the site level. TIMES-NZ 3.0 model files and input data are openly available at github.com/EECA-NZ/TIMES-NZ-Model-Files and documented at times-nz-dev.readthedocs.io.
Mathematical structure. TIMES is a linear programme. The objective function minimises the discounted total system cost over the planning horizon. Constraints encompass technology capacity balances, resource availability limits, demand satisfaction requirements, and policy mandates including the coal phase-out NES and the ETS carbon price trajectory. The LP structure assumes all relationships between decision variables and costs are linear, which is a strong simplification that enables tractable optimisation over a 30 to 50 year horizon at national scale but forecloses the representation of non-linear grid dynamics, threshold effects, and the temporal coincidences that determine hosting capacity exceedances.
Technology representation. Each energy conversion technology is represented as a process with declared capital cost, operating cost, thermal efficiency, and asset lifetime. The investment decision for each technology in each period is an endogenous model variable: the optimiser decides how much capacity to build, when to build it, and when to retire it. This produces an economy-wide investment pathway rather than a single-year snapshot, and the technology mix in any given year is determined by the full inter-temporal optimisation rather than by year-by-year myopic decisions.
Temporal Resolution: The Critical Limitation
TIMES-NZ 2.0 represented a full year using 24 time slices — combinations of season and time-of-day that together approximate the variety of supply and demand conditions across the year. The temporal resolution of TIMES-NZ 3.0 has not been publicly confirmed in the model’s documentation at the time of writing and should be verified against times-nz-dev.readthedocs.io before the 24-slice characterisation is assumed to carry over. Regardless of any improvement in temporal resolution in the 3.0 rebuild, this is adequate for capturing seasonal variation in renewable generation (winter hydro scarcity versus summer hydro abundance) and broad demand patterns, but it is entirely inadequate for capturing the hour-by-hour grid dynamics that determine whether a large industrial electrification commitment triggers a GXP hosting capacity exceedance.
Subsequent NZ-specific implementations have partially addressed this. A more recent NZIES model instance uses 96 time slices (24 hours multiplied by 4 seasons), requiring generation and transfer capacity to meet demand in both seasonal and daily peak periods. Even at 96 slices, however, the model cannot represent the specific coincidence of events that constitutes a GXP exceedance: the winter peak hour when Edendale’s electric boilers are drawing maximum load, a dry hydro year has reduced South Island generation headroom, and competing industrial users in the Mararoa-Waimea zone are simultaneously at high demand. This coincidence is an hourly event requiring the 8,760-timestep representation that the DCM framework’s DemandPack and SignalsPack architecture provides.
This structural temporal gap is the primary reason why TIMES-NZ cannot perform the analysis that Module 6 performs. The 23-of-64 GXP exceedance finding is not visible in a 96-slice or 24-slice national model. It only emerges when hourly demand traces are evaluated against hourly grid signal data across a structured future ensemble — precisely the architecture this framework provides.
Spatial Resolution: Two Regions
TIMES-NZ 2.0 aggregated New Zealand to two regions — broadly North Island and South Island — connected by the HVDC inter-island link. The spatial resolution of TIMES-NZ 3.0 has not been publicly confirmed at the time of writing and should be verified at times-nz-dev.readthedocs.io. This captures the fundamental geographic constraint of the NZ electricity system without representing the distribution-level constraints within each island that determine site-level feasibility.
The two-region structure cannot distinguish Southland from Canterbury. It cannot represent the Mararoa-Waimea transmission cluster. It cannot assess whether the Fonterra Edendale facility and Alliance Group’s Lorneville processing plant, both in eastern Southland, are competing for the same zone substation headroom. These spatial distinctions are precisely what determines the infrastructure feasibility of the investment decisions that EECA’s GIDI programme is designed to accelerate. The gap between two-region national representation and GXP-level site representation is the spatial dimension of the analytical gap that the DCM framework fills.
Sensitivity Analysis: Scenario Comparison versus DMDU Ensemble
TIMES-NZ has reached a third major iteration (3.0), which represents a complete rebuild rather than an incremental update. The two TIMES-NZ 3.0 scenarios are named Steady and Shift — they are designed to help users explore structural trade-offs and uncertainties, not to provide probability-weighted predictions or prescriptive pathways. (TIMES-NZ 2.0 used scenarios named Tūī and Kea; these have been superseded.)
TIMES-NZ 3.0 sensitivity analysis is conducted through comparison of exactly two named scenarios: Steady and Shift. EECA explicitly describes them as instruments for exploring data rather than a comprehensive coverage of plausible futures. The two scenarios span four structural uncertainties identified by EECA as critical:
Table SM-5.5-A-TNZMAPPING: TIMES-NZ 3.0 critical uncertainties mapped to DCM FutureArtefact dimensions
| TIMES-NZ 3.0 critical uncertainty | Steady assumption | Shift assumption | DCM FutureArtefact dimension |
|---|---|---|---|
| Economic structure | Primary sector and traditional manufacturing continues | Structural shift away from primary exports | D_demand_mult — heat demand multiplier |
| Global technology progress | Moderate clean technology cost reduction | Rapid cost reduction for wind, solar, battery, EVs | U_upgrade_capex_mult — capital cost trajectory |
| Individualistic vs cooperative | ETS carbon price $52/tonne by 2035 | ETS carbon price $260/tonne by 2050 | P_ETS — ETS price multiplier |
| Role of gas | LNG import via floating terminal viable | Gas exits energy mix; no LNG import | New structural branch: LNG access binary |
The gap between two nationally consistent structural scenarios and a 64-to-1,000 point DMDU ensemble is wider than any scenario count comparison implies. EECA’s own framing acknowledges it: two scenarios are a starting point for conversation, not a coverage of the uncertainty space. The Steady and Shift scenarios do, however, provide precisely calibrated anchor points for the FutureArtefact ETS dimension: $52/tonne (Steady, 2035) as the lower bound and $260/tonne (Shift, 2050) as the upper bound. These publicly attributed, citable figures from EECA replace any previously generic language about carbon price trajectories in the DCM ensemble calibration.
The Steady/Shift distinction is directly Southland-relevant: the Steady scenario describes dairy and meat processing as ‘a steady source of income with international farming exports finding growing markets,’ while the Shift scenario describes ‘demand for energy used by meat and dairy processing plants falling because of lower international demand.’ The two scenarios bracket exactly the range of futures that the Edendale PoC must navigate. This is not an abstract macroeconomic parameter — it is the primary structural driver of site-level heat demand uncertainty in the Southland context.
The strengths of this approach are transparency and interpretability: each scenario is a coherent, self-consistent account of a possible future. The limitation is coverage. With only two national structural scenarios in TIMES-NZ 3.0, the analysis explores only a tiny fraction of the combinatorial uncertainty space, and the scenarios are designed as anchor points for conversation rather than to identify the specific combinations of conditions that most strongly determine pathway preference — which is precisely what PRIM-based scenario discovery of the type specified in Sub-Module SM-2.4-A is designed to reveal.
The RETA Southland report explicitly recommends that future RETA pathways make greater use of sensitivity analysis to illustrate how resource costs and carbon prices influence the pathway choice, acknowledging that current sensitivity methods are insufficient for communicating the depth of genuine uncertainty about which pathway is preferred. The DCM framework’s DMDU ensemble approach is the direct methodological response to that recommendation: where TIMES-NZ 3.0 produces two internally consistent national structural scenarios, the DCM ensemble evaluates 64 to 1,000 structured combinations of the uncertain drivers that most influence the site-level decision.
Integration Pathway: TIMES-NZ and the DCM Framework
The most defensible integration approach uses TIMES-NZ 3.0 scenario outputs to calibrate the boundary conditions of the FutureArtefact ensemble rather than to constrain the site-level decision directly.
One-directional calibration (current implementation, §5.6). TIMES-NZ 3.0 scenario outputs provide plausible national electricity price trajectories, concrete carbon price anchors ($52/tonne in Steady by 2035; $260/tonne in Shift by 2050), and sectoral demand projections including dairy and meat processing trajectories under different technology adoption assumptions. These outputs are translated into the upper and lower bounds of the FutureArtefact’s ETS price dimension and the electricity tariff multiplier range, ensuring that the futures explored in the DCM ensemble are consistent with plausible national energy trajectories rather than with arbitrary parameter ranges constructed independently of the national system context.
LNG binary structural branch. TIMES-NZ 3.0 introduces a new structural uncertainty not present in 2.0: whether New Zealand develops LNG import infrastructure via a floating terminal (Steady scenario) or exits gas entirely (Shift scenario). For Southland industrial heat users, this binary materially affects the counterfactual fuel price floor: if LNG import is viable, the economics of electrification relative to gas retention are different from those in a gas-exit world. The FutureArtefact ensemble should include this as a structural branch parameter — a discrete binary rather than a continuous multiplier — in future ensemble designs that extend beyond the current PoC’s biomass versus electricity framing.
Open-data calibration pathway. TIMES-NZ 3.0 model files and input data are publicly available at github.com/EECA-NZ/TIMES-NZ-Model-Files under an open licence. The one-directional calibration step described above can therefore be implemented with full reproducibility using public data, without any data-sharing arrangement. This directly satisfies the ReferenceArtefact schema’s requirement for a stable, publicly accessible source with a maintained URL. The documentation site at times-nz-dev.readthedocs.io provides the versioned specification needed for the ReferenceArtefact’s calibration_method field.
Gauss-Seidel iterative coupling (next-phase vision, §7.2). The aggregate industrial pathway commitments revealed by DCM ensemble analysis — for example, that 60 percent of plausible futures prefer biomass over electrification for the Southland dairy cluster — can be summarised as industrial demand scenarios and passed back to TIMES-NZ as updated demand constraints. TIMES-NZ 3.0 re-runs under those constraints, producing updated national electricity price and carbon price trajectories, which recalibrate the FutureArtefact ensemble for the next iteration. Convergence is achieved when site-level and national-level projections are mutually consistent. This coupling architecture is the operational expression of the multi-scale analytical environment described in Module 7 §7.2, and it is technically feasible with existing tools.
The Wider NZ Energy Modelling Ecosystem
Table SM-5.5-A: NZ energy modelling tools — capabilities and DCM integration roles
| Tool | Type | Temporal res. | Spatial res. | Primary NZ role | DCM integration role |
|---|---|---|---|---|---|
| TIMES-NZ 3.0 | LP optimisation | 24 slices/year | 2 regions | National least-cost pathways | FutureArtefact ETS and price calibration |
| NZIES | LP optimisation | 96 slices/year | 3 to 5 regions | Hydrogen and deep electrification analysis | Enhanced calibration with HVDC constraints |
| PyPSA-NZ | Network LP | Hourly (8,760) | Nodal | Electricity dispatch and investment | Regional Module implementation (SM-6.6-E) |
| REMix-NZ | Integrated simulation | Hourly | Sub-regional | Renewable integration and multi-carrier demand | Regional electricity calibration; demand projection |
| RETA methodology | Screening | Annual and peak | GXP-level | Industrial decarbonisation opportunity mapping | SignalsPack headroom calibration; SM-5.2-A |
| GIDI Fund analysis | Project appraisal | Project lifetime | Site-level | Co-investment eligibility and grant assessment | Site-perspective cost adjustment in overlay |
| EnergyPLAN | Simulation | Hourly | National | Renewable energy system simulation | Alternative Regional Module implementation |
| TIMES (IEA-ETSAP) | LP optimisation | Configurable | Configurable | Global and regional energy pathways | Methodological parent of TIMES-NZ |
PyPSA-NZ is the most important next-phase Regional Module implementation target. It provides nodal network optimisation at hourly resolution for the New Zealand transmission system, producing dual variables and line loading fractions that map directly to SignalsPack fields: the nodal marginal price at EDN0331 becomes the tariff_nzd_per_mwh field; the line loading fraction on the subtransmission circuit becomes the feasibility_indicator; the system adequacy margin becomes headroom_mw. The national PyPSA-NZ model can be subset to the Southland regional context, reducing computational cost while retaining network-aware dispatch accuracy. Full implementation is specified in Sub-Module SM-6.6-E.
REMix-NZ, developed at DLR Stuttgart with NZ researchers, provides integrated energy system modelling with hourly resolution across electricity, heat, and transport carriers. Its published demand projection dataset — released in 2026 with regional and hourly resolution to 2050 — provides exactly the calibration data required for the DCM framework’s regional demand growth multiplier dimension, covering sector-specific electrification trajectories at sub-national spatial resolution that TIMES-NZ cannot provide. The full dataset, providing hourly and regional energy demand projections for electricity, heat, and transport in New Zealand to 2050, was published as an open dataset in January 2026 with DOI 10.1038/s41597-025-06511-6 (Canessa et al., 2026), directly providing the calibration data required for the DCM framework’s regional demand growth multiplier dimension.
The ReferenceArtefact: Governed External Document Linkage
The DCM framework’s analytical backbone currently traces lineage from DecisionSummaryArtefacts through ResultArtefacts, SignalsPacks, FutureArtefacts, and DemandPacks back to generating runs and module versions. One important link in this chain is not yet formally governed: the relationship between the FutureArtefact’s uncertain driver ranges and the external documents — RETA reports, Transpower Grid Investment Plans, Climate Change Commission modelling, and TIMES-NZ scenario outputs — from which those ranges were calibrated.
The ReferenceArtefact closes this gap. It is a new canonical artefact family that registers external documents as governed analytical inputs with full provenance. A ReferenceArtefact carries: the document’s title and institution; its URL or DOI; its publication date; the specific FutureArtefact fields or module sections it calibrates; the calibration methodology used to translate document content into parameter ranges; and a SHA256 hash of the source file where a downloadable version exists.
Once registered, every FutureArtefact field has a traceable chain: the ETS price range in a given future references the ReferenceArtefact for the Climate Change Commission’s Demonstration Path modelling, which in turn references the specific table or figure from which the price bounds were derived. The SignalsPack headroom estimate references the ReferenceArtefact for the RETA Southland assessment, linking the analytical finding back to the EECA report that provided the reference capacity data. This makes the complete audit trail from framework results back to primary data sources recoverable without analyst interpretation.
Table SM-5.5-A-2: ReferenceArtefact schema
| Field | Type | Required | Description |
|---|---|---|---|
| ref_id | UUID | Required | Unique identifier for this reference |
| title | String | Required | Full document title |
| institution | String | Required | Authoring institution |
| authors | String | Optional | Primary authors or report number |
| url_or_doi | String | Required | Stable URL or DOI |
| publication_date | Date | Required | Date of publication or last revision |
| calibrates_fields | List | Required | FutureArtefact fields this document calibrates |
| calibrates_sections | List | Optional | Module section numbers this document informs |
| calibration_method | String | Required | How document content was translated into parameter ranges |
| file_sha256 | String | Optional | SHA256 hash of downloaded source file |
| schema_version | String | Required | “1.0” |
Table SM-5.5-A-3: Current ReferenceArtefact registry entries (planned)
| Document | Institution | Calibrates | Method |
|---|---|---|---|
| RETA Southland Assessment | EECA | headroom_mult range | Reference capacity from GXP screening table |
| Climate Change Commission — Demonstration Path | Climate Change Commission | ets_price range | Table 5.3 NZU price trajectory low and high bounds |
| TIMES-NZ 3.0 Energy Scenarios | EECA / BEC | Steady and Shift scenario outputs; ETS price anchors ($52/t and $260/t); sectoral demand projections | github.com/EECA-NZ/TIMES-NZ-Model-Files; times-nz-dev.readthedocs.io |
| TIMES-NZ 3.0 Documentation Site | EECA | Versioned model specification for calibration_method field in ReferenceArtefact | times-nz-dev.readthedocs.io |
| EECA TIMES-NZ Model Files | EECA-NZ (GitHub) | TIMES-NZ model inputs and scenario data | Direct download from github.com/EECA-NZ/TIMES-NZ-Model-Files |
| Transpower GIP 2024 | Transpower | upgrade_capex_mult range | Indicative subtransmission upgrade cost schedule |
| PowerNet PF-025 | PowerNet Limited | Customer Contribution methodology | Section 5.4 subtransmission capacity trigger thresholds |
| REMix-NZ demand projections | DLR Stuttgart / University of Auckland | demand_growth_mult range | Regional industrial electricity demand 2030–2050 |
This specification is designated as a next-phase backbone extension. The schema is declared here; implementation follows the governed backbone development pathway specified in Sub-Module SM-3.7-A.
Future Extension: The Knowledge Context Layer
Beyond governing links to external documents, the ReferenceArtefact registry provides the foundation for a Knowledge Context Layer — a structured manifest that an AI agent or analyst can query to determine which external sources are relevant to any given FutureArtefact field, module section, or analytical finding. When the natural-language query interface described in Module 7 §7.4 is deployed, the Knowledge Context Layer is what allows it to ground its answers in specific, citable primary sources rather than in the model’s training data.
The Knowledge Context Layer also supports the tagging system of §7.8: a section tagged [Research] can carry a ReferenceArtefact link to the specific literature gap it identifies, making the open research question not only declared but traceable to its empirical foundation. A section tagged [Collaborative] can carry a ReferenceArtefact link to a comparable implementation in another domain, lowering the barrier for domain specialists who want to contribute an instantiation.
This layer is declared here as a Vision component. Its specification will be developed as a separate sub-module in Module 3 (Architecture) once the primary backbone implementation described in SM-3.7-A is complete.
[Collaborative]
This sub-module invites contributions in three areas: calibration data for TIMES-NZ scenarios, PyPSA-NZ implementation for the Regional Module, and additional ReferenceArtefact registry entries linking external NZ energy analyses to specific FutureArtefact fields.
End of SM-5.5-A