Backtest

To assess the accuracy of our model, we use realized, historical data as inputs and compare the output prices, to histroical real-time prices.

Backtest input data

In the backtest outline below the following historical data is used as input data.

InputHistorical data setMethodology
Energy demand60d SCED Generation Resource Disclosure, Hourly load archiveCalculated as sum of SCED base points across all technologies. Zonal split informed by hourly load archive.
Ancillary service demand2-Day Ancillary Services ReportsTotal ancillary service demand is the total of all self-arranged and cleared volume across all ancillary services.
Available generation capacity60d SCED Generation Resource DisclosureFor each technology type and zone, available generation capacity is calculated as the daily maximum HASL (high ancillary service limit) of online units (those with a telemetered resource status other than OUT).
Load factors60d SCED Generation Resource DisclosureFor each technology type and zone, load factors are calculated as the ratio of 15-min HSL (high sustained limits)
Gas priceHenry Hub Natural Gas Spot PriceHenry Hub spot prices is used for historical gas pricing without adjustment (other than appropriate unit conversions).
Coal price60d SCED Generation Resource DisclosureVolume-weighted median three-part offer bid prices from units with a resource status of CLLIG excluding those of lignite units.

Backtest results

For the following graphics we compare: 15-minute, realized, real-time bus average (HB_BUSAVG) pricing; and 15-minute, load-weighted average of zonal prices.

Error distribution

Sample price comparisons