microgrid

microgrid models a solar + battery plant and produces marginal energy prices: what one more kWh costs the site, slot by slot over the next two days, and what a kWh is worth by source. A flexible load, or a scheduler for one, reads those prices to decide when to run.

It is a Python library: it takes the plant's description and its inputs as time series.

flowchart LR
    subgraph in["Inputs (time series, from any source)"]
        prices["import and export prices"]
        pv["PV forecast<br/>(quantiles)"]
        load["load forecast"]
        soc["battery SOC,<br/>meters"]
    end
    plant["plant/2<br/>the plant's topology"] --> engine
    prices --> engine["microgrid.engine"]
    pv --> engine
    load --> engine
    soc --> engine
    engine --> out["per slot: marginal cost, regime,<br/>surplus, certainties,<br/>values by source, supply stack"]

plant/2 is the plant's description (reference); each output is defined in Outputs.

How it prices

• It prices the next kWh, not the average one. A kWh used in a slot comes from one source: PV that would have been exported, the grid, or the battery. The price is that source's: the export given up, the import, or the battery's energy. The price follows the plant forward to find out which, slot by slot. The methodology explains how.
• It knows what the battery holds. A kWh from the battery costs what its energy cost: the day's solar at the export it gave up, the night's grid energy at what it was bought for, through the battery's losses and wear. And a full battery with sun ahead is priced to be used: kept full, it would waste the day's PV.
• It prices the weather's uncertainty. The price is an average over PV scenarios, published with a pessimistic case and the chance it holds.
• It degrades conservatively. A missing or stale input moves the price toward the import price and says why. A stale discount is never reused.

Where to start

• Try it. Install, then run your first valuation on the example dataset: one synthetic week of an example plant, no hardware needed.
• Understand it. The methodology, from "why one more kWh" to what a battery kWh costs and the PV scenarios.
• Use it on your plant. Describe your plant, build a dataset from your own records, or call it from Python.
• Look something up. The reference: the formats, the microgrid command and the API.
• A term you don't know (SOC, DC bus, hybrid inverter, ...): the glossary.

The parts

Part	What
microgrid.plant	the plant model: the plant described as data (plant/2) and projected slot by slot
microgrid.engine	the price model, the battery's lots, the series and units its inputs arrive in, the checks against the meters
microgrid.cli	the microgrid command: the library run on a dataset (dataset/1) of files, or on a site's live sources (sources/1; the sources extra).
examples/dataset	one synthetic week of an example plant, with the outputs the repository's tests check the command against.