The Business Strategy Assessment Model (BSAM) is an agent-based electricity wholesale market simulation model which simulates the complex operations within a power pool, central dispatch, day ahead electricity market. The model simulates electricity generators as entities who progressively learn to bid their capacities in the wholesale electricity market, with ultimate goal the maximization of their profits. In parallel, a unit commitment and optimal dispatch algorithm calculates the quantities injected by each generation unit, the system marginal price, the system costs, as well as derived outputs, such as, CO2 emissions and profits of each generator. The model can support cost-benefit analysis of future policy and/or technology deployment scenarios. It is very flexible since it simulates the power generators as agents that compete with each other and adapt to policy and/or market changes.
Key Features
The novelty of BSAM lies in:
- Combining a reinforcement learning approach to facilitate market participant decision making and bidding behaviour, with Unit Commitment and Economic Dispatch algorithms, used to accurately model a power-pool central dispatch electricity market at the day-ahead level. This novel integration of Unit Commitment optimization with dynamic agent-based market participation, is highly suitable for the investigation of the market effects of a wide variety of issues, including highly-uncertain capacity development scenarios, high RES penetration, and agent behaviour changes under the context of complex market transitions.
- Featuring modularity, flexibility, soft-linking capabilities and good computational performance. The above features present a comparative advantage compared to typical unit commitment algorithms in electricity systems with low residual demand levels, whose computational cost is high and their integration capability limited. The good performance of BSAM is an ideal fit for modelling complex, mid- and long-term scenarios. Indicatively, BSAM requires on average 12–14 seconds per day modelled.
BSAM is developed using exclusively the Python language.
Main Modules
BSAM consists of three modules: (i) a wholesale electricity market module, (ii) an agent-based module, and (iii) a unit commitment module.
Wholesale electricity market module: This module is used to simulate the operation of the wholesale market by appropriately managing market-related data. More specifically, within this module, the net electricity demand to be met is calculated, the spinning reserve requirements are applied, low utilised power plants are decommissioned, and profiteering agents are fined.
Agent-based module: This module simulates the bidding behaviour of electricity generating resources (as self-learning profit-maximizing agents) participating in a Day-Ahead competitive wholesale electricity market.
Unit commitment module: This module simulates a central dispatch day-ahead wholesale electricity market clearing procedure and outputs the respective generating resources’ dispatch schedule, as well as the system marginal price for each trading period within the 24-hour time horizon of the following day. The simulations aim to maximize social welfare, considering the agents’ electricity bids and the generating resources’ constraints.
Impact
The BSAM model has been developed and applied in the context of the following projects funded by the European Commission:
Scientific Articles and Other Relevant Publications
- Kontochristopoulos, Y., Michas, S., Kleanthis, N., & Flamos, A. (2021). Investigating the market effects of increased RES penetration with BSAM: A wholesale electricity market simulator. Energy Reports, 7, 4905-4929. https://doi.org/10.1016/j.egyr.2021.07.052
- Nikas, A., Stavrakas, V., Arsenopoulos, A., Doukas, H., Antosiewicz, M., Witajewski-Baltvilks, J., & Flamos, A. (2020). Barriers to and consequences of a solar-based energy transition in Greece. Environmental Innovation and Societal Transitions, 35, 383-399. https://doi.org/10.1016/j.eist.2018.12.004
Video Presentation
A study performed with BSAM, soft-linked with the STREEM and AIM models, is presented in the video below.