August 13th, 2020
The primary function of a production cost model is to determine which generators must be committed and dispatched during each interval of the simulation and the associated cost of operation. Simulation intervals are typically 1 hour (as performed in this analysis), but there is increased interest in sub-hourly simulations, especially in scenarios of increased VG penetration where sub-hourly net load variability can require increased dispatch flexibility (CAISO 2011). A simple dispatch is determined by “stacking” generators in order of production cost (from lowest to highest) until the sum of the individual generator output is equal to load in each time interval. The actual dispatch is complicated by the many additional constraints imposed by individual generators, such as minimum up and down times and ramp rates. The actual dispatch also depends on the need for system security, including spinning reserves, which consist of partially loaded generators with the ability to rapidly ramp in response to a generator outage or unexpected increase in demand. To determine the optimal dispatch requires detailed information for each generator. Primary characteristics include maximum capacity, minimum stable output level, plant heat rate (ideally as a function of load), fuel cost, ramp rates, start time, and minimum up and down time. The software then co-optimizes the need for energy and reserves subject to the various constraints and finds the least-cost mix of generators in each time interval.
VG plants with little or no variable cost are typically placed into production cost models as a fixed hourly generation profile. Because they have no variable cost, and may also have production incentives, they are typically dispatched first but may be curtailed when operational con
straints do not allow the system to accept their output. These constraints might occur when the VG exceeds the local capacity of the transmission network or during periods when conventional generators have reduced output to their minimum generation levels. This second phenomenon can be referred to as a “minimum generation” problem (Rogers et al. 2010) or an “over generation” problem (CAISO 2010). CSP plants without storage can be placed into the model in the same manner as a wind or PV plant, using the hourly output from a CSP simulation model.
For this study, trough CSP plants (both with and without storage) and PV were simulated using the System Advisor Model (SAM) (Gilman et al. 2008; Gilman and Dobos 2012) version 2012-5-11. The CSP simulations used the wet-cooled empirical trough model (Wagner and Gilman
. The model converts hourly irradiance and meteorological data into thermal energy and then models the flow of thermal energy through the various system components, such as losses in the HTF, finally converting the thermal energy into net electrical generation output. The CSP plant without storage assumes a solar multiple of 1.3, the SEGS VIII default power block with turbine over-design operation allowed at 105% and used default settings for all parameters, such as parasitics. Meteorological data was derived from the National Solar Radiation Database from 2006 (NREL 2007).
CSP with TES was implemented in this study using a two-step process. First, hourly electrical energy from the CSP plant was simulated using SAM, and then the electrical energy was dispatched in PLEXOS using a combination of algorithms that largely existed within that model. This process is illustrated in Figure 3 and described in more detail in the following sections.
The first step (hourly electrical energy) was created using SAM in a manner similar to the case without storage with several important differences. Essentially all parameters that are affected by plant dispatch were moved out of SAM and into the PLEXOS framework. First, the solar multiple was set to 1; a larger solar multiple and storage was implemented in the PLEXOS model as described later in this section. Second, the minimum generation levels and start-up energy requirements were set to 0 and also accounted for in PLEXOS. Parasitics were removed from the gross CSP generation to derive a net hourly generation. Operational parasitics calculated by SAM were subtracted from the electrical profile in a manner similar to other thermal power stations. We also considered the constant parasitic loads (e. g., associated with fluid pumps) that occur even when the plant is not operating. This means that the CSP plant will draw a small amount of energy from the grid and incur a small associated cost. This constant load was calculated separately based on SAM CSP simulations.
FIGURE 4: The flow of energy through a trough CSP plant with TES in PLEXOS
The product of the SAM simulation is “raw” electrical energy output, which is then processed in PLEXOS using a modified form of the PLEXOS hydro algorithm to simulate storage, generator operation, and the effect of an oversized solar field. In each hour, the model can send the electrical energy from the SAM simulations directly to the grid via a simulated power block, to storage, or a combination of both. The model can also choose to draw energy from storage. The simulated power block includes the essential parameters of the CSP power block, including start-up energy, minimum generation level, and ramp-rate constraints. The model considers start-up losses in the dispatch decision by assuming a certain amount of energy (equivalent to 20 MWhe for a 100-MW power block) is lost in the start-up process.
In CSP plants that use indirect storage, the additional efficiency losses in the storage process are also simulated. The storage losses are set to 7%, which capture both the efficiency losses in the heat exchangers and the longer-term decay losses. Constant parasitics were added by placing a constant load on the same bus as the CSP plant. The general implementation is illustrated in Figure 4, representing a 200-MW CSP plant with a solar multiple of 2.0 and 6 hours of storage.
Figure 4 also shows the effect of solar multiple, which is captured in the sizing of the power block and storage components. For example, a solar multiple of 2.0 can be simulated by setting the maximum size of the power block to 50% of the maximum output from the CSP simulations from SAM. Likewise, the storage system can be sized to accommodate some fraction of the maximum CSP output. The storage energy capacity (hours of storage) can be set independently.