-
Notifications
You must be signed in to change notification settings - Fork 5
Methodology
Many descriptions of the internal operation of the NEM reference concepts of the bid stack or merit order. This explanation of dispatch expresses clearly the intent of the NEM to preferentially dispatch low cost generation and high value load. Further to this description the caveat, that the dispatch process takes place subject to relevant constraints is often added. In the more detail descriptions the co-optimisation energy and FCAS markets is also discussed. These standard descriptions of the NEM are mentioned here to highlight the divergence of the description made below, not just in the level of detail, but in the focus. The implementation of the merit order will be outlined, but a large portion of the detail needed to describe the NEM in fact pertains to the mechanisms that constrain its operation and couple the provision of services across both markets and regions.
The simplest high level view of the internal operation of Nemlite (and by design the NEM) consists of three concepts, the set of system variables, the objective function and the constraint matrix, together these form a linear program. The system variables represent optimisable quantities for any given dispatch. The objective function describes the total operating cost of the market, it is linear and its assigns each system variable a cost coefficient. The constraint matrix consists of rows that are linear inequalities, these can describe simple bounds on single system variables or create relationships between large sets of systems variable. Describing the market in this way allows for the application of a linear solver to minimise the cost of the system subject to an arbitrary number of linear constraints.
The concept of the bid stack can be mapped directly to the underlying implementation, the system variables represent how much of each component of the bid stack is dispatched, the cost coefficient of each variable represents the price of each bid and the constraint matrix places bounds on each variable equal to their bid quantity.
While frame work of a linear program easily implements the concept of a bid stack, it also contains the potential to represent the variety of other properties the NEM needs to function, including distribution loss factors, regional nodes, interconnectors with variable losses, ramp rates, flexible constraints and multiple coupled service markets.
| System Variables | x1 | x2 | x2 | ... | xn | ||
|---|---|---|---|---|---|---|---|
| Objective Function | c1x1 | c2x2 | c2x2 | ... | cnxn | = | C |
| Constraint matrix | a1,1x1 | a1,2x1,2 | a1,2x2 | ... | a1,nxn | => | b1 |
| . | |||||||
| am,1x1 | am,2x2 | am,2x2 | ... | am,nxn | => | bm |
In the modelling of the NEM dispatch process the construction of the constraint matrix is the most complicated task. Here the method used in Nemlite is broken down into steps that each describe the process of representing a particular aspect of the NEM inside the constraint martix.
Each dispatch unit in the NEM submits can submit up to 10 bids for each service and for each dispatch interval, these are then implemented as system variable with the optimisation process determining how of each bid should be dispatched. The following sections outline the hard bounds placed on these variables by the constraint matrix.
The bid quantity is a particular upper bound placed on each system variable that represents a bid. For each bid in the market there is a corresponding constraint of the form:
an,mxn <= bm
Where xn is the bid system variable, bm is the bid quantity and the coefficient anm is always equal to 1.
Dispatch units also submit a maximum availability as part of their bid offer which places a bound on the dispatch of their combined bids. For each set of bids for a particular service there is a corresponding constraint of the form:
an,mxn + ... + an+9,mxn+9 <= bm
Where xn to xn+9 are the 10 system variables associated with a bid set, bm is the offered availability and the coefficients an,m to an+9,m are each always equal to 1.
Additionally to the offered availability another availability measure that includes variable generation's forecast availability is used. The resultant constraint is of the same form as above but with a differing right hand side value, hence only the most restrictive of the two constraints will be binding. An extra caveat is that while the offered availability exists for both energy and FCAS bids, this second availability exists only for energy bids.
Ramp rates place both an upper and lower bound on the sum of each set of energy bids. The upper and lower bounds have corresponding constraints of the form:
an,mxn + ... + an+9,mxn+9 <= bm
an,m+1xn + ... + an+9,m+1xn+9 >= bm+1
Where xn to xn+9 are the 10 system variables associated with a bid set. The constant bm is the generators initial operating level at the start of the dispatch interval plus the lesser of the offered and telemetered ramp up rates. The constant bm+1 is the generators initial operating level at the start of the dispatch interval minus the lesser of the offered and telemetered ramp down rates. The coefficients an,m to an+9,m and an,m+1 to an+9,m+1 are each always equal to 1.
To ensure that the co-dispatch of energy and FCAS markets is technically feasible a series of constraints that couple a dispatch unit's dispatch into the separate markets is needed.
The concept of the FCAS trapezium has previously been described in detail by AEMO (AEMO, 2015). In brief it allows all dispatch units to define their own set of constraints that couple a set of FCAS bids to their set of energy bids. This is done by letting each dispatch unit offer 4 additional values with their FCAS bids, these are an enablement minimum value, an enablement maximum value, a low break point and a high break point. These together with the units total FCAS availability are used to define a set of three inequalities that confine a units dispatch to a trapezium shaped zone on the plane defined by units energy dispatch on the y-axis and the units FCAS dispatch (for a particular service) on the x-axis. These constraints can be thought of as the trapeziums lower slope, upper slope and . . .