Our power market services utilise our proprietary National Electricity Market Renewable Energy Simulator (NEMRES), which calculates annual capacity changes, energy generation, and transmission expansion decisions based on intra-hourly dispatch, imitating AEMO’s dispatch engine.
Various rules, laws and policies govern the operation of the NEM, with the key elements being power supply almost always matching power demand, adjusted for constraints in the transmission and distribution network, and energy storage. The supply side is comprised of demand side participation, energy storage, fossil fuel, and renewable generators that offer capacity based on calibration with current offers and dispatched from the least to more costly offers, subject to system conditions, to meet demand.
Demand is affected by several factors such as weather, economic activity, population, etc. Although demand for power has patterns, it is generally unplanned and highly inelastic. System operators rely on demand forecasting for the daily market operation and long-term planning. As such AEMO publishes forecast demand over different time frames regularly, which we apply based on our analysis of annual investments.
NEMRES simulates the NEM least cost dispatch processes to match supply and demand conditions in the forecast periods, modelling the resulting energy generation and emissions from each of scheduled and semi-scheduled plant.
The figure below outlines the main model components and model process flows. The central component of NEMRES is the least cost dispatch model, which dispatches the generation of plants based on default bids calibrated to each generator’s most recently observed patterns.
Merit Order Model
A merit order is constructed via the bids offered by all scheduled and semi-scheduled plants. Our algorithm orders the price bands offered by plants from the least to highest and accumulates the quantities of corresponding power offers accordingly.
For each dispatch interval, default bids are optimised for individual facility profitability. Hydro generation is allocated by the model based on current dam levels and the associated capacity of ‘run-of-river’ generation (intra-daily and weekly reservoirs) and storable hydro capacity (inter-annul and seasonal). As shown in the Figure above, the input data preparation and model calibration are important blocks, supported by several criteria in checking the validity of model outputs for capacity expansion modelling. This includes analyst checks against closing facilities based on model profitability, and the feasibility of new entrants in each region based on publicly announced projects, transmission constraints, and zonal build limits.
The default bidding model constructs four price and quantity pairs. These pairs reflect the percentage of the estimated costs and available capacity of each plant. Estimates are standardised based on historic operational patterns and calibrated based on the latest available market data.
The first price band of a default bid applies to generation that would incur higher costs for being dispatched down – e.g. turned-off. The second band relates to the short-run marginal cost (SRMC) or the fuel cost multiplied by the heat rate plus the variable operating cost for existing and committed facilities. For example, renewable facilities normally have a SRMC less than $10 per Megawatt-hour (MWh), while coal-fired generators fall between $10 and $50 per MWh and gas-fired plants are greater than $60 per MWh. The third offer relates to a Levelised Cost of Energy or long-run marginal cost (LRMC) target. The last band is affected by the facility’s competition within their region.
The quantity pair is the percentage that a plant is willing to offer to the market at the four offers outlined above. The quantity is incremental, in that the sum of the four quantity components must be 100 per cent. Contracts impact the percentage of electricity subject to bidding behaviour and spot price revenue. The quantity at the SRMC cost is related to the generator’s estimated contract level, while the quantity at the LRMC is allocated to the normal design level less the amount that has already been allocated in the previous price bands. The last band can be normally be thought of as quality held back to maximise profit.
Although the model can operate across a range of time-steps, our outlooks typically hybridise two bidding formats. Long-term forecasting calculates dispatch on annual load duration curve and is used for inter-annual forecasting. Half-hour (or shorter), high precision modelling is performed on critical periods to better resolve which facilities are dispatched in medium-term situations, e.g. the next 16 quarters.
The cost of a generator depends on several factors: plant characteristics such as generator type, region, maximum capacity, summer and non-summer ratings, outage and repairs rates for both full and partial unavailability, maintenance durations, minimum loads, minimum and maximum capacity factors, fixed and variable operating costs, heat rates, fuel costs, emissions, marginal loss factors, auxiliary loads, short-run marginal costs, and expected retirement date, etc. Of these variables, fuel costs are normally updated quarterly, whereas most other variables are usually adjusted annually.
Offer strategies are adjusted based on plant profitability. Annual and/or quarterly profit is calculated as total revenue from the sent-out energy plus any fixed subsidies less the variable cost associated with per MWh generation and less the annual fixed cost.
Annual forecast demand comes with three predictions for the NEM. One is for annual energy consumption and the other two are for maximum and minimum demand loads. RepuTex fits historical demand profiles to AEMO’s various forecasts and aims to mimic the modelling intervals between 48 to 17,520 periods per year, equivalent to averaging demand over 7.60 to 0.02 days. Leap days, public holidays and weekend load profiles are checked and matched as required.