(485a) Multiscale Optimization of Integrated Energy Systems Using Machine Learning Models for Market Interactions | AIChE

(485a) Multiscale Optimization of Integrated Energy Systems Using Machine Learning Models for Market Interactions


Ghouse, J. - Presenter, McMaster University
Jalving, J., Sandia National Laboratories
Knueven, B., Sandia National Laboratories
Siirola, J., Sandia National Laboratories
Miller, D., National Energy Technology Laboratory
Dowling, A., University of Notre Dame
Gao, X., University of Notre Dame
Chen, X., Carnegie Mellon University
Agi, D., University of Notre Dame
Achieving deep decarbonization of the power sector requires large-scale adoption of renewable energy sources. New sources of flexibility from both consumers (demand) and generators (supply) are critical to offsetting the variability for non-dispatchable renewable sources and maintaining the continuous balance of the electric grid. Integrated Energy Systems (IES) [1] can provide this flexibility by synchronizing material and energy flows to produce multiple valuable products such as electricity, heat, steam, or chemicals in hybrid configurations, ultimately enabling greater new renewable integration, more curtailment of emissions, and reductions in grid operating costs [2]. A key advantage afforded by IES flexibility is the means to participate in electricity markets by offering energy and regulation services. Because an IES can dynamically apportion its energy output (e.g., using thermal and/or electrical storage), it can provide responsive generation to grid operators and help to improve grid reliability [3]. Consequently, the design and optimization of an IES must consider both the technical performance of the energy system and its economic viability within dynamic energy markets.

Fundamental challenges toward the development of an IES include accounting for multiple energy market time scales (e.g. real-time, day-ahead, and long-term capacity markets) [4], and exogenous uncertainty that arises due to decisions as a market participant that affect the entire system. The effects of exogenous uncertainty can be particularly important to capture. For instance, an IES design decision like the size of storage will impact how it bids into the market (e.g., ramping rate), which impacts both (1) the market revenues of the entire system and (2) how the IES is asked to respond from the system operator. Market prices are also often set by the marginal resource, i.e., the resource with the highest cost out of those that are dispatched. As a result, slight perturbations in the parameters of an IES that is marginal (such as generation) can induce large fluctuations in the market, both for the IES and its neighbors. Accepted modeling approaches typically use "price-taker" assumptions [5] that do not capture the exogenous uncertainty inherent to designing new generation systems, which necessitates the development of new optimization formulations.

In this talk, we propose novel computational approaches to co-optimize the design and operation of IES while explicitly capturing market interactions (i.e., exogenous uncertainty). We start by conducting over 64,000 annual simulations of the RTS-GMLC system while varying the characteristics of a single generator using the open-source Prescient cost production model [6-8]. Here, each simulation corresponds to replacing a single generator in the network with a hypothetical IES. Sensitivity analysis [9,10] confirms that generator parameters such as the marginal production and startup costs are most influential on market outcomes. From this large simulation library, we develop algebraic [11] and neural network surrogate models to predict the impact of these generator characteristics, i.e., market inputs, on market outcomes including generator revenue and the dispatch profile. We then formulate mixed-integer and/or nonlinear two-stage stochastic programs to co-optimize the design (e.g., sizing of components, stage 1 decisions) and operations (e.g., power generation profile, stage 2 decisions) of an IES. Prescient simulations of the optimized IES design are used to compare the predicted versus actual market outcomes such as revenue and capacity factors. Using simple IES examples, we show how our proposed surrogate approach outperforms the de facto standard price taker approach which ignores market interactions. We also discuss extensions to multi-period problems with dynamic surrogates. We conclude by highlighting the importance of incorporating complex IES/market interactions, especially when considering future scenarios such as deep carbonization where energy infrastructure is more tightly coupled.


[1] M. O. Malley et al., “Energy Systems Integration : Defining and Describing the Value Proposition”, 2016

[2] D. Arent et al., "Multi-input, Multi-output Hybrid Energy Systems," Joule, 2021

[3] A. W. Dowling and V. M. Zavala, “Economic opportunities for industrial systems from frequency regulation markets,” Computers and Chemical Engineering, 2018

[4] F. Sorourifar, V. M. Zavala, and A. W. Dowling, “Integrated Multiscale Design , Market Participation , and Replacement Strategies for Battery Energy Storage Systems,” IEEE Transactions on Sustainable Energy, 2020

[5] A. W. Dowling, R. Kumar, and V. M. Zavala, “A multi-scale optimization framework for electricity market participation,” Applied Energy, 2016

[6] B. Knueven, J. Ostrowski, J. P. Watson, "On Mixed-Integer Programming Formulations for the Unit Commitment Problem", INFORMS Journal on Computing, 2020

[7] J. P. Watson, B. Knueven, R. Concepcion, D. Melander, A. Short, P. Zhang, D. Woodruff, Prescient, Computer software, Available: https://github.com/grid-parity-exchange/Prescient

[8] C. Barrows et al., “The IEEE Reliability Test System: A Proposed 2019 Update,” IEEE Transactions on Power Systems, 2020

[9] E. Plischke, E. Borgonovo, and C. L. Smith, “Global sensitivity measures from given data,” European Journal of Operational Research, 2013

[10] A. Saltelli, P. Annoni, I. Azzini, F. Campolongo, M. Ratto, and S. Tarantola, “Variance based sensitivity analysis of model output . Design and estimator for the total sensitivity index,” Computer Physics Communications, 2010

[11] A. Cozad, N. V Sahinidis, and D. C. Miller, “Learning Surrogate Models for Simulation-Based Optimization,” AIChE Journal, 2014


This work was conducted as part of the Design Integration and Synthesis Platform to Advance Tightly Coupled Hybrid Energy Systems (DISPATCHES) project through the Grid Modernization Lab Consortium with funding from the U.S. Department of Energy’s Office of Fossil Energy and Carbon Management, Office of Nuclear Energy, and Hydrogen and Fuel Cell Technology Office. Additional work was conducted as part of the Institute for the Design of Advanced Energy Systems (IDAES) with support through the Simulation-Based Engineering, Crosscutting Research Program within the U.S. Department of Energy’s Office of Fossil Energy and Carbon Management. Additional work was conducted in part by appointments to the U.S. Department of Energy (DOE) Postgraduate and Faculty Research Programs at the National Energy Technology Laboratory administered by the Oak Ridge Institute for Science and Education (ORISE).

This work was authored in part by the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-NA-0003525. This work describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government.

This work was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.