Bali

Bali is the demo built on a real client inventory: the south of the island — Denpasar, the Pesanggaran power complex, Ngurah Rai airport and the port of Benoa — from two spreadsheets received from the operator, coordinates in UTM. The domain was deliberately scoped to southern Bali, and eleven facilities outside the grid (Celukan Bawang, Pemaron, Gilimanuk…) were left out: what is modelled is one power plant and its neighbourhood, not the whole island.

Time zone	Asia/Makassar
Standard evaluated	Indonesia, Government Regulation PP 22/2021, Annex VII (id-pp22-2021)
Grid	100 × 100 cells of 750 m, about 75 × 75 km
Forecast	108 hours, every 15 minutes
Sources	8 in 3 groups: 7 stacks and 1 area
Receptors	8, sensitive points declared by the operator

The receptors

There is no monitoring network here: the eight receptors are the sensitive points the operator marked on a map — two schools, two transport terminals (the airport and the Benoa cruise-ship terminal) and four tourist sites (the marine park, Waterbom, the Eco Tent Kura camp, the Kertalangu cultural village). The two schools are the receptors closest to the complex — the State Senior High School sits a few hundred metres north of the engine stacks — and are the first to cross a level. A forecast here is not compared with measurements; it is compared with itself across scenarios.

The sources, in three groups

Pesanggaran Power Complex — twelve Wärtsilä 50DF engines rated 17.6 MW each plus two boilers, around 200 MW nominal, the largest engine-based power plant in Indonesia. It runs mainly on LNG regasified at the Tanjung Benoa terminal and can switch to liquid fuel. The twelve engines are modelled as four stacks of three, and dispatch is expressed by scaling a single engine’s rate. It is the domain’s dominant source of nitrogen oxides, and being dual-fuel is what makes it worth more than one scenario: an interruption at a single regasification terminal changes the emission profile of the whole domain without anything failing.

Denpasar Waste-to-Energy — one stack. Designed for 1,000 to 1,500 tonnes of waste a day, under construction in 2026, commercial operation expected in 2028. It is modelled in every scenario regardless, because the question this demo answers is what the airshed looks like once the facility runs.

Fugitive dust — the plant’s waste pile as an area source, PM10 only. The group exists to keep that difference legible in a contribution breakdown; its single rate is the least well-founded number in the inventory.

The scenarios

Partial Dispatch on Gas is the base case: four of the twelve engines and both boilers on natural gas, SO₂ negligible, an NOₓ-dominated profile.

LNG Supply Interruption — the engines switch to liquid fuel. Designed plant behaviour rather than an upset; the change is not uniform: SO₂ multiplies by more than four hundred, NO₂ by four, particulate by three. The plume changes composition rather than simply growing. Receptor alerts are evaluated against this scenario: NO₂ reaches 198.8 % of the PP 22/2021 limit at the worst point.

Full Plant Dispatch — all twelve engines running, on the same fuel: the engine stacks are scaled by three and everything else stays as base. It changes how many units run, not what they burn, and separates the two questions a viewer asks separately: how much is emitted, and what is emitted.

The alerts

Alerts are evaluated against the Indonesian standard PP 22/2021, with the averaging period each species requires. A receptor crossing a level shows on the map and in the alerts panel; past and forecast are told apart by their position against the forecast cycle.

A note on the figures

The engine rates are screening values from AP-42 factors and the manufacturer’s specifications — total NOₓ rather than NO₂, and a particulate figure dominated by the condensable fraction. The comparison with the operator’s continuous monitoring is still open. They serve to compare scenarios with one another; they are not permit figures.

→ The demos · Air-quality standards