Economic Assessment and Design of Concentrated Solar Power Systems
Concentrated solar power (CSP) systems with thermal energy storage (TES) are promising technologies to provide renewable, on-demand electricity. These systems are also capable of providing ancillary services (regulation, spinning reserves, etc.), which are required to stabilize the electricity grid and mitigate variations in demand and generation from intermittent renewable sources. Most techno-economic analyses of CSP technologies, however, focus on Levelized Cost of Electricity (LCOE) which neglects the time-varying value of electricity and supplemental revenues from ancillary services. For example, a CSP plant providing 10 MW of regulation capacity (an ancillary service) for all hours of 2015 in the California energy market would have received $500,000 in capacity payments alone. In fact, some studies have demonstrated divergent conclusions regarding the profitability of CSP designs using LCOE versus other economic metrics (Return on Investment, Net Present Value, etc.).
Revenue estimation, which is required for these alternate economic metrics, may be formulated as a large-scale optimization problem: given market price and solar radiation data, determine the operating policy (e.g., mass and energy flows) that maximizes revenues. Considering an entire year with detailed physics-based models (e.g., mass and energy balances, nonlinear Rankine cycle efficiency correlations, etc.) and start-up/shut-down restrictions results in a non-convex mixed integer nonlinear program with millions of continuous variables and tens of thousands of binary decisions. As such, most studies rely on model simplifications to reduce complexity. In this project, we seek to develop a unified framework for CSP economic analysis with the following features:
1. Evaluate revenues from selling/providing both energy and ancillary services, including regulation and (non)-spinning reserve capacity, in day-ahead and real-time markets
2. Use a full year or statistically meaningful representative sample of historical coincident meteorological and electricity/ancillary service market price data
3. Consider detailed (nonlinear) performance models for steam cycle, solar collector and TES, as well as start-up/shut-down limitations (e.g., minimum up/down times) and efficiencies
We are specifically interested in the application of massively parallel solvers for structured mathematical programming, such as PIPS-NLP and DSP, and physics-inspired relaxations to make CSP revenue estimation computational tractable. We seek to optimize CSP designs with respect to multiple economic and performance metrics (objectives), including LCOE, NPV and fresh water usage for a variety of technologies. The methodologies we are developing are applicable to a large class of energy systems, for which operational policy and design decisions are tightly coupled and must be co-optimized.