Statistical Learning for Optimal Energy Market Participation Under Uncertainty

Institute for Design of Advanced Energy Systems (IDAES) Grid Modeling

Grid operations incorporate decision-making processes on time scales covering 12 orders of magnitude:

Through the IDAES project, we are developing multiscale modeling capabilities that explicitly integrate individual resource operational decisions and market clearing. This enables unique analysis capabilities:

• Elucidate complex relationships between resource dynamics and market dispatch (with uncertainty, beyond price-taker assumption)
  • Compute expected revenues
  • Understand how resource bids impact market dispatch and prices
• Predict the economic opportunities and market impacts of emerging technologies (e.g., H₂ production, integrated energy systems)
  • Quantify the economic benefit of system flexibility 
  • Compare novel system concepts under varying market scenarios
• Guide conceptual design/retrofit to meet current and future power grid needs
  • Balance market revenue and long-term equipment health
  • Properly size hybrid systems/storage considering both detailed dynamics and sub-hourly markets

Through the Design Integration and Synthesis Platform to Advance Tightly Coupled Hybrid Energy Systems (DISPATCHES), we demonstrate and quantify the benefits of potential hybrid systems based on three case studies:

• Coal (existing and Coal FIRST) with energy storage, etc.
• Existing nuclear with hydrogen production and utilization, etc. 
• Renewables with dispatchable options (batteries, NGCC)

Multiscale Optimization of Integrated Energy Systems and Market Interactions

Accelerating the deep decarbonization of the world’s electric grids requires the coordination of complex energy systems and infrastructures across timescales from seconds to decades. We develop a new multiscale simulation framework that integrates process- and grid-centric modeling paradigms to better design, operate, and control integrated energy systems (IESs), which combine multiple technologies, in wholesale energy markets. Traditionally, IESs are analyzed with a process-centric paradigm such as levelized cost of electricity (LCOE) or annualized net revenue, ignoring important interactions with electricity markets. Our framework explicitly models the complex interactions between an IES’s bidding, scheduling, and control decisions and the energy market’s clearing and settlement processes, while incorporating operational uncertainties. 

 

IES Optimization with Machine Learning Surrogates to Capture Market Interactions

 

Co-optimizing the Design and Operation Strategy of Solid Oxide Fuel Cell-Based Electricity-Hydrogen Coproduction Systems 

The rapid adoption of non-dispatchable renewable energy increases electricity market volatility and creates an urgent need for more flexible energy systems to balance supply and demand. Integrated energy systems (IES) can offer this flexibility by combining multiple technologies and providing the option to switch between multiple inputs and outputs (e.g. electricity + hydrogen).  Detailed market analysis of such IES systems is a challenging task, due to nonlinear models of system operations and modeling switching between operation modes.We develop a framework for rapidly evaluating IES concepts through optimization-based market-informed techno-economic analysis (TEA). Detailed equation-oriented process models are developed in the IDAES® PSE modeling platform. We then use ALAMO to generate algebraic surrogates for operating costs, capital costs, and co-production constraints.  These surrogates enable us to account for complex system dynamics in larger, time-dependent models. Finally, these surrogates are embedded in a Generalized Disjunctive Programming (GDP) model to account for mode-switching system behaviors and the GDP model is solved to output optimal size and output schedule. Here, we model seven IES concepts which hybridize Natural Gas Combined Cycles, SOFC, SOECs, rSOFCs, and compressed air energy storage (CAES) across over ten future LMP scenarios under varying carbon tax policies and VRE penetration levels. This will ultimately develop guidance on which strategies to incorporate SOFC and SOEC technologies into the grid are most promising. 

 

Probabilistic Forecasts for Multiscale Energy Prices

Gaussian Process (GP) has been widely used for regression and classification. An autoregressive method with GP regression for energy prices has been tailored for Day-ahead Market (DAM). The immediate prices are used as the inputs and historical price data is used to train the model. GP regression interfaces decision-making under uncertainty and energy market bidding naturally. One of the ongoing works is to develop forecasting techniques for faster market layers (e.g. Fifteen-minute Market (FMM)) combining the information gained in slower layers (DAM).

Related Publications 

Xian Gao, Bernard Knueven, John D. Siirola, David C. Miller, Alexander W. Dowling (2022). Multiscale Simulation of Integrated Energy System and Electricity Market Interactions. Applied Energy, 316, p. 11901. [link]

Nicole Cortes, Xian Gao, Bernard Knueven, and Alexander W. Dowling. Estimating Energy Market Schedules using Historical Price Data (2022). 14th International Symposium on Process Systems Engineering (PSE2021+). Ed. by Y. Yamashita, M. Kano. [file]

Clay T. Elmore, Alexander W. Dowling (2021). Learning Spatiotemporal Dynamics in Wholesale Energy Markets with Dynamic Mode Decomposition, Energy, 232, p. 121013 [preprint] [link]

Andrew Lee, Jaffer H. Ghouse, John C. Eslick, Carl D. Laird, John D. Siirola, Miguel A. Zamarripa, Dan Gunter, John H. Shinn, Alexander W. Dowling, Debangsu Bhattacharyya, Lorenz T. Biegler,  Anthony P. Burgard, David C. Miller (2021). The IDAES Process Modeling Framework and Model Library – Flexibility for Process Simulation and Optimization, J. Advanced Manufacturing and Processing, 3(3), p. e10095. [link]

Farshud Sorourifar, Victor M. Zavala, Alexander W. Dowling (2020). Integrated Multiscale Design, Market Participation, and Replacement Strategies for Battery Energy Storage Systems, IEEE Trans. Sustainable Energy, 11(1), p. 84-92. [link]

Xian Gao and Alexander W. Dowling. Making Money in Energy Markets: Probabilistic Forecasting and Stochastic Programming Paradigms. 2020 American Control Conference (ACC), Denver, CO, USA, 2020, p. 168-173. [link] [video] [file]

Alexander W. Dowling, Tian Zheng, and Victor M. Zavala (2018). A decomposition algorithm for simultaneous scheduling and control of CSP systems, AIChE Journal 64 (8), p. 2480-2417. [link]

Alexander W. Dowling and Victor M. Zavala (2018). Economic Opportunities for Industrial Systems from Frequency Regulation Markets, Computers & Chemical Engineering 114 (9), p. 254-264. [link]

Alexander W. Dowling, Tian Zheng, and Victor M. Zavala (2017). Economic Assessment of Concentrated Solar Power Technologies: A Review, Renewable and Sustainable Energy Reviews 72, p. 1019–1032. [link] [preprint]

Alexander W. Dowling, Ranjeet Kumar, Victor M. Zavala (2017). A Multi-Scale Optimization Framework for Electricity Market Participation, Applied Energy 190, p. 147-164. [link] [preprint]