(177g) A Novel Framework for Optimal Combined Capacity Planning and High-Resolution Scheduling of Renewable Energy Systems

Renewable energy systems must be cost competitive for widespread adoption, so determining economically optimal renewable energy system configurations and unit capacities is essential. The time-varying nature of intermittent renewables means that these decisions cannot be made for a single steady-state or a small discrete set of operating points. The prevailing approach in literature is the explicit inclusion of scheduling decisions at the capacity planning stage, yielding optimal combined capacity planning and scheduling (OCCPS) models [1-3]. These models consist of unit selection and capacity decisions made once for a given system, as well as unit scheduling decisions (e.g. on/off, production rate, storage inventory) made for each period of a scheduling horizon which captures both diurnal and seasonal variation in renewable availability. These “on/off” or unit commitment (UC) decisions are formulated as binary variables, introducing a combinatorial complexity generally considered to make the use of a full year, hourly resolution scheduling horizon (8760 periods) intractable [4,5]. In previous research, the size of the scheduling decision space has been reduced via temporal aggregation of renewable generation and power demand data into representative periods, where authors have considered, for example, 72 days with hourly resolution (1728 periods) [1], an hourly resolution week per season (672 periods) [2], or 672 variable length multi-hour periods determined via clustering of consecutive hours [3].

In this work we propose a new iterative algorithmic framework for OCCPS of energy systems which makes tractable full year hourly resolution problems and even allows for sub-hourly scheduling decision resolutions. This new framework lessens the computational burden caused by binary UC decisions through selective iterative enforcement of the UC constraints. Specifically, in renewable energy systems, many candidate process units have a wide partial load range and can be cycled on/off quickly. For example, PEM electrolysis can operate at partial loads as low as 5% of installed capacity and can dynamically cycle in seconds [1]. We exploit these characteristics by initially relaxing “fast cycle” units’ UC constraints in the OCCPS problem, significantly reducing the number of binary decisions. We iteratively solve relaxed problems, identifying which relaxed UC constraints are violated and enforcing only these constraints in the next problem iteration. Selective iterative enforcement continues until all UC lower bounds are satisfied. Benders decomposition is used to accelerate the solution of each iteration to enable overall convergence in reasonable time. We systematically determine the Benders master- and sub-problems by using core-periphery identification [6] on the variable graph of the OCCPS problem, where variables appear as nodes which are connected by edges if they share common constraints.

We demonstrate our proposed OCCPS framework using a case study based on the University of Minnesota Morris (UMM) campus. The campus currently meets power demand through a combination of two 1.65 MW local wind turbines without any energy storage and purchases from a utility under demand charge tariff structure, whereby in addition to quantity charges, the highest 15-minute interval of power demand each month is charged at a demand rate. We pursue OCCPS of a local energy storage system which includes battery, hydrogen and ammonia candidate energy storage pathways. The objective of the OCCPS problem is to minimize the cost of power supply to the Morris campus, using local energy storage to mitigate demand charges. The demand charge structure necessitates scheduling at a 15-minute resolution, giving 35040 operating periods; an OCCPS problem of this size is not tractable with a standard solver. We demonstrate the tractability of this problem using our proposed framework and highlight the cost savings enabled by making scheduling decisions at such a granular resolution.

References

[1] Gabrielli, P., Gazzani, M., Martelli, E., & Mazzotti, M. (2018). Optimal design of multi-energy systems with seasonal storage. Appl. Energy, 219, 408-424.

[2] Zhang, Q., Martín, M., & Grossmann, I. E. (2019). Integrated design and operation of renewables-based fuels and power production networks. Computers & Chemical Engineering, 122, 80-92.

[3] Palys, M. J., & Daoutidis, P. (2020). Using hydrogen and ammonia for renewable energy storage: A geographically comprehensive techno-economic study. Comput. Chem. Eng, 127, 106785.

[4] Goderbauer, S., Comis, M., & Willamowski, F. J. The synthesis problem of decentralized energy systems is strongly NP-hard. Computers & Chemical Engineering 2019, 124, 343-349.

[5] Sass, S., Faulwasser, T., Hollermann, D. E., Kappatou, C. D., Sauer, D., Schütz, T., Shu, C. Y., Bardow, A., Gröll, L., Hagenmeyer, V., Müller, D., & Mitsos, A. (2020). Model compendium, data, and optimization benchmarks for sector-coupled energy systems. Computers & Chemical Engineering, 135, 106760.

[6] Zhang, X., Martin, T., & Newman, M. E. (2015). Identification of core-periphery structure in networks. Phys. Rev. E, 91(3), 032803.