Methodology

Production forecasting definitions, formulas, validation design, and interpretation safeguards

Methodology

Distributed Electricity Definition

Distributed electricity is the annual electricity distributed to customers within a harmonized modeling region, reported in gigawatt-hours (GWh). Historical values cover 2011–2024. Production forecasts for 2025–2030 use the regional Drift model.

Local Generation Definition

Local generation is the annual electricity generated within the same harmonized region, reported in GWh. It does not represent total electricity available to the region because interregional imports, exports, and power-system flows are outside the model target. Production forecasts use Drift V1.

Production Forecast Formula

Average Annual Change
slope = (y_T − y_1) / (t_T − t_1)
Forecast
ŷ_(T+h) = y_T + h × slope

Drift estimates the average annual change implied by the first and latest usable observations in each regional series, then extends that change into the forecast horizon. The calculation is applied separately to every harmonized region and separately to distributed electricity and local generation.

Active Model Variables

Distributed electricity target
Annual distributed electricity (GWh)
Local generation target
Annual local generation (GWh)
Time index
t = annual observation year
Exogenous predictors (X)
None in production
Model class
Univariate annual time-series models

Population, GRDP, inflation, technology mix, and other enriched variables were assessed in challenger experiments. They are not inputs to the active production forecasts.

Production Forecasting Approach

Historical period
2011–2024
Forecast horizon
2025–2030
Modeling geography
34 harmonized regions
Frequency
Annual
Distributed electricity model
Drift
Local generation model
Drift V1
Regional uncertainty intervals
Not published

Validation Design

Rolling One-Step
Target years
2019–2024
Forecast horizon
1 year
Purpose
Near-term stability
Recursive Multi-Step
Origins
2018–2021
Maximum horizon
3 years
Purpose
Forecast-path robustness

Supported evaluation measures are macro sMAPE, macro MASE, micro RMSE, aggregate bias, and challenger win rate against the Drift benchmark.

Accounting Difference Formula

Accounting Difference (GWh)
Local Generation − Distributed Electricity
Accounting Difference (%)
Accounting Difference / Distributed Electricity × 100
Local Generation Ratio
Local Generation / Distributed Electricity

Positive and negative values are spatial accounting differences. They do not represent operational surplus, shortage, reserve margin, or blackout probability.

Accounting Difference Categories

Far Above Distributed Electricity
≥ +20%
Above Distributed Electricity
≥ +5% and < +20%
Relatively Balanced
> −5% and < +5%
Below Distributed Electricity
> −20% and ≤ −5%
Far Below Distributed Electricity
≤ −20%
Unavailable
Incomplete pair

These boundaries follow the production classification exactly. Values equal to +5% are classified as Above, while values equal to −5% are classified as Below.

Harmonized Geography

The dashboard uses 34 harmonized modeling regions. Papua Selatan, Papua Tengah, and Papua Pegunungan are represented within Papua, while Papua Barat Daya is represented within Papua Barat. Earlier Kalimantan Timur and Kalimantan Utara observations follow the audited geography-break treatment. Values are not duplicated across split provinces.

Historical Data Treatment

  • • The 2016 historical distributed-electricity and local-generation pairs include audited interpolation for continuity. They are explicitly identified as interpolated and are not treated as observed values.
  • • Nine historical pairs remain incomplete: DI Yogyakarta in 2011–2014 and 2020, plus Kalimantan Timur and Kalimantan Utara in 2011–2012.
  • • Missing historical values are preserved and are never converted to zero.
  • • Forecast years 2025–2030 contain 204 complete region-year pairs.

Interpretation Caveat

Important: The regional accounting difference is a spatial accounting proxy. Electricity may flow across interconnected regions; therefore, local generation below distributed electricity does not imply shortage or blackout risk, while local generation above distributed electricity does not imply freely available operational capacity. System-level utility planning data are required for adequacy, reliability, reserve margin, and investment decisions.