4.2 Post unit commitment relaxation (PUCR) with interleaving
4.2.1 Interleaved model simulation
The interleaving tool’s main function is to account for the assumption of perfect foresight in the day-ahead (DA) scheduling of generation. The models in Chapter 7 and Chapter 8 were developed with both a DA and a real-time (RT) model interleaved with each other.
The interleaved model is based on the running of the SEM, which is scheduled on a DA basis for the next day with the assumptions shown in Table 4.3 and the optimisation of each day is divided into intra-day trading periods [12], as shown in Table 4.1. To account for the effects of wind energy forecasting errors and unforeseen forced outages, the simulations use two models running interleaved with each other. The first is a DA model which is followed by a RT model. Having the models running linked to each other helps to achieve accurate results for the simulation of dispatch of generators on the AI electricity system. The DA and RT models pass information back and forth between each other at the end of every simulation day as illustrated in Fig. 4.2.1. This information sent from the DA model to the RT model at the end of the DA model’s simulation day includes the DA UC, generation, and interconnector flow schedules. At the end of the RT model simulation day the RT model sends the generators’ initial conditions back to the DA model to start the simulation of the next simulation day. The DA and RT models are represented in Eqn. 4.1.
4. METHODOLOGY: PLEXOS
IMPLEMENTATION
4.2 Post unit commitment relaxation (PUCR) with interleaving
Accounting for the forecast error in the day-ahead unit commitment (DA UC) schedule leads to differences in generation cost as well as differences in generation dispatch due to varying degrees of accuracy of the DA schedules being sent to the RT model.
Table 4.1: SEM optimisation time-line
Time Event
06.00hr D Schedule commences
48 (30 minute) 48 (30 minute) interval optimisation 05.30hr D+1 Schedule ends
06.00hr D+1 Lookahead period for model optimisation begins 6 (1 hour) 6 (1 hour) interval optimisation
11.30hr D+1 Lookahead period for model optimisation ends
The DA model optimises on a short term schedule as shown in Table 4.1. The function of the DA model is the creation of DA UC schedules for all large generators in the AI system, which are listed in Table 4.2. The DA model also creates the DA interconnector flow schedules and fixed generation schedules.
The DA model is optimised based on the DA wind forecast data. Scheduling of the DA model is carried out stochastically to account for the known wind forecast uncertainty, this is described in Section 4.3. The DA model includes maintenance in the simulation as the maintenance schedule is known in advance of the DA UC schedule. The DA model however does not include forced outages as these in reality will occur randomly and without advance knowledge.
The RT model re-optimises the final schedule using the DA schedules.
However, these are subject to constraints and post unit commitment relaxation (PUCR) which are discussed in depth in Section 4.2.2. The RT model optimises on the same schedule as the DA model as described in Table 4.1. The RT model includes forced outages and the same maintenance
schedule as the DA model. The same maintenance and forced outage file is used across all scenarios and is developed from the base-case scenario of the particular simulation.
Investigation of factors driving the costs of operating the 2020 Irish power system with large-scale wind generation.
43 Edward V. Mc Garrigle
Figure 4.2: The PLEXOS optimisation process showing information used in the process represented by “AI system optimisation”
Figure 4.3: Flowcart representing the simulation steps of the DA and RT mod-els’ interleaved process
4. METHODOLOGY: PLEXOS
IMPLEMENTATION
4.2 Post unit commitment relaxation (PUCR) with interleaving
Figure 4.4: Flowcart representing the simulation steps of the DA and RT mod-els’ interleaved process
Investigation of factors driving the costs of operating the 2020 Irish power system with large-scale wind generation.
45 Edward V. Mc Garrigle
Generator name Generator ID ROI
Aghada Unit 1 AD1
Aghada Unit 2 ADC
Dublin Bay Power DB1
Great Island GI CCGT (new)
Huntstown Phase 2, HN2
Huntstown Phase 1 HNC
Moneypoint Unit 1 MP1
Moneypoint Unit 2 MP2
Moneypoint Unit 3 MP3
Northwall Unit 4 NW4
Poolbeg PBC
Tynagh TY
Whitegate WG
NI
Ballylumford Unit 10 B10 Ballylumford Unit 31 B31 Ballylumford Unit 32 B32
Kilroot Unit 1 K1
Kilroot Unit 2 K2
The daily solutions of the DA and RT models may be described by the following Equations:
DA(d − 1|d) = f (W F (d), ICinf o, M odels,inf o(d)) where d = 1
DA(d − 1|d) = f (W F (d), ICinf o, RT.Endsyst,con(d − 1), M odels,inf o(d))
where d = 2, 3, 4, . . . , n
(4.1)
RT (d|d) = f (W A(d), DA.U Cpucr(d), DA.ICf ix(d), DA.Genf ix(d), F O(d), M odels,inf o(d))
where d = 1, 2, 3, . . . , n (4.2)
M odels,inf o(d) = Sysdemand(d), Fcost,
OPconst, Genconst, GBsytem,inf o(d),
M aint(d) (4.3)
4. METHODOLOGY: PLEXOS
IMPLEMENTATION
4.2 Post unit commitment relaxation (PUCR) with interleaving
where: M odels,inf o(d)= system information given to both DA and RT models;
M aint(d)= maintenance schedule set for both DA and RT models;
DA(d − 1|d)= DA model solution for day d on d − 1; RT (d|d) = RT model solution for day d on d; f (W F (d)) = DA wind energy forecast and uncertainty quantiles time-series; f (W A(d)) = realised actual wind energy time-series;
ICinf o= interconnector characteristics; RT.Endsyst,con = end system conditions from the RT model for the DA initial conditions; DA.U Cpucr(d)= DA UC schedule with post unit commitment relax; DA.ICf ix(d)= fixed
interconnector flow schedule from DA model; DA.Genf ix(d)= fixed generator flows schedule on day d from DA model (hydro, waste, biomass and CHP units); F O(d) = forced outages; Sysdemand(d)= system demand; OPconst= operational constraints; Genconst= generator profile constraints, ensuring minimum capacity factors and reducing ramp cycling (Hydro, Waste,
Biomass, Peat and CHP units); Fcost= Fuel costs; GBsystem inf o= Great Britain wind generation, system demand and price settings; d = day interval in 2020;
n= 366 days in 2020.