Thermal energy storage (TES) systems increase profitability of gas-turbine-powered generating plants by exploiting daily pricing patterns to chill water in off-peak hours and then provide turbine inlet-air cooling in peak demand periods to boost output and improve heat rate.
However, TES systems often are operated using default running schedules based on vendor design calculations, which do not account for the actual prices of electricity and natural gas, or weather conditions, or the thermodynamic state of the plant. The high variability in external and internal conditions during plant operation implies that a fixed operating policy is sub-optimal in real-world situations.
This means further significant benefits—perhaps $1500 to $3000 per day for 500-750-MW combined cycles—may be available by optimizing TES operation. Real Time Power Inc (RTP) shared with attendees at the recent Combined Cycle Users Group annual meeting (Orlando, Aug 24-27, 2015) the company’s automated software solution for computing the optimal TES run schedule for both day-ahead and real-time markets.
RTP’s David Davis explained the application and how data from both the powerplant and energy trading floor are used in state-of-the-art optimization techniques to maximize cash flow within a given time period.
A typical TES system produces chilled water which can be stored in a large well-insulated tank or sent directly to cooling coils in the gas-turbine air inlet house (Fig 1). Inlet air also can be cooled using chilled water from the tank, a preferred option for peak demand hours to reduce parasitic power consumption.
The chilled and unchilled water are stored in the same tank; there is little mixing between the two layers. The thermocline level is defined as the height in the tank where there is greatest temperature difference between the warm water above and the cool water below.
TES optimization is performed by maximizing the cash flow derived from the chilled-water resource while respecting all of the physical and operational constraints. Silvia Magrelli, the R&D software engineer at RTP who led development of the application said the optimal schedule requires the “solution of a constrained multivariable nonlinear objective function, and with the selection of the appropriate solver algorithm, the machine-generated answer will provide significant additional revenue gain over and above the default schedule.”
The optimizer is layered on high-accuracy adaptive plant models which allow it to calculate key powerplant operational parameters for the forecasted weather conditions, and then also to predict precisely the incremental power and heat-rate benefit of inlet-air cooling, as well as the auxiliary load of the chiller units.
The final pieces of the jigsaw puzzle are forecasts of electricity and natural-gas prices, which enable decisions regarding GT run schedule. The outcome assures sufficient chilled water is available to meet the forecasted needs of the day-ahead market. Plus, in real-time trading it allows sudden changes in market and/or weather conditions to be assimilated immediately and the best use of the TES re-computed for the remainder of the trading day.
One important benefit of the optimizer, Team RTP stressed: It greatly improves the accuracy of day-ahead load forecasts because it always plans to return the TES tank to a specified state at the end of each day. Thus, at the start of each trading day, the thermocline level is the expected value, and the forecasted load and heat rate can be achieved if the TES is operated according to the optimal schedule.
By contrast, with a default schedule, the end-of-day thermocline level is uncontrolled and will affect megawatt and heat-rate numbers for the following day. Finally, when plant equipment constraints are reached—such as maximum chilled-water flow rate through the cooling coil—the system will plan the day-ahead and real-time schedules based on best achievable performance.
Real Time Power has tested the application against five years of real price and weather data using actual plant thermodynamic models. Figs 2-4 presents the results of this retrospective comparison for a 3 × 1 combined cycle in the South with a 5.75-million-gal storage tank. The optimizer consistently outperforms the default schedule, and over the five-year period investigated, would have produced a $6.6-million benefit. The saving results from differences between the actual electricity price on the day and price curve used in designing the TES and in compiling the default schedule.
Installation of the RTP solution involves a server which connects to both the plant DCS and the energy trading desk. Control-room operators have the opportunity to set independent variable values and constraints—gas-turbine availability, for example. The system typically is accessed daily by the trader to produce the day-ahead declaration, and subsequently, as required during the current day, to make changes reflecting the real-time market.
The optimal solution for a particular running configuration is computed in less than one second, making the system very responsive to changes in market and weather conditions.