Internal Colloquium - Comparative Simulation of Energy Market Designs

Florian Maurer

May 6, 2024

Outline

Motivation
Research Question
Requirements
Functional Requirements
Non-Functional Requirements
Summary
Project Plan
Publication list

Motivation

Easy way to compare market designs

Motivation

Easy way to compare market designs
Comparative evaluation with simulation

Motivation

Easy way to compare market designs
Comparative evaluation with simulation
Like PyPSA but for energy markets

Research question

How can different energy market designs be evaluated in a simulative comparative manner?

Objective: Framework for the comparison of market designs

Which abstract description is able to characterize different market designs?

In what way can different market mechanisms be represented in the simulation software?

How can key performance indicators of market mechanisms be used to evaluate specific market designs?

Market comparability

Comparison agent-based market simulation

Model	Year	Market Design	OSS	Power Flow	Modularity	Interoperability	Scaled	Market Comparison
GSY-e [1]	2016-now	LEM	✓	○	✓	✓	✓	○
lemlab [2]	2021-now	LEM	✓	○	○	◐	◐	○
SIMONA [3]	2021-now	○	✓	✓	○	◐	○	○
ASAM [4]	2021-now	DA, ID, RE	✓	◐	○	○	○	○
AMIRIS [5]	2016-now	DA	✓	○	◐	○	◐	○
USEF [6]	2016-2017	USEF	✓	◐	○	○	○	○
FNCS/AMES [7]	2008-now	SCUC/SCED	✓	◐	✓	◐	◐	○
Powermatcher	2014-2017	○	◐	○	○	◐	✓	○
PowerACE [8]	2013-2016	DA	◐	○	○	○	○	○
Maon [9]	2012-2021	DA	○	✓	○	○	○	○
ÖkoFlex [10]	2014-2017	DA	○	◐	○	○	○	○

table from [11]

○ - not applicable, ◐ - partial applicable, ✓ - fully applicable

Missing features of current market simulations

Features
various configurable market designs
OTC & conditional contracts
concurrent auctions
grid optimization
foundation for reinforcement learning
comparative evaluation

Summary of requirements

Functional

Generic market design abstraction
Modelling of different market mechanisms
Possible grid optimization through power flow calculation
Customizable output metrics
Evaluation using key performance indicators

Non-Functional

Availability/Open Source
Modularity
Interoperability with existing tools
Scalability

Functional requirements

Generic market design abstraction
Modelling of different market mechanisms
Possible grid optimization through power flow calculation
Customizable output metrics
Evaluation using key performance indicators

First research question

Which abstract description is able to characterize different market designs?

Market abstraction

market config item	description
name	string name
product type	energy or capacity or heat
market products	list of available products to be traded
opening hours	recurrence rule of openings (RFC 5545)
opening duration	time delta
market mechanism	name of method used for clearing
maximum bid	max allowed bidding price
minimum bid	min allowed bidding price
maximum volume	the largest valid volume for a single bid
additional fields	list of additional fields to base bid
volume tick size	step increments of volume
price tick size	step increments of price
volume unit	string for visualization
price unit	string for visualization
supports get unmatched	boolean
maximum gradient	max allowed change between bids
eligible obligations lambda	function checking if agent is allowed to trade here

concept presented in [11]

Market product

Market product	Example
delivery duration	1 hour
count	24
offset to opening	24 hours

Base bid

Contains fields which are mandatory
Common abstraction layer
Can be extended by market
Additional fields receive meaning through market mechanism

field	value
start-delivery	2023-06-30 00:00
end-delivery	2023-06-30 01:00
volume	100
price	32
sender id	118

Example for Germany - Day-Ahead Auction[12]

Example for Germany - Intraday Auction

Example for Germany - Intraday Trading

Example for Germany - Futures

First research question

Which abstract description is able to characterize different market designs?

Abstract description created
Presented in OS2 and EI.A paper [11]
Fits to various market designs

First research question

Which abstract description is able to characterize different market designs?

Abstract description created
Presented in OS2 and EI.A paper [11]
Fits to various market designs

Second research question

In what way can different market mechanisms be represented in the simulation software?

Agent-based simulation approach

All agents have
- Obligations
- Message handlers
Event-based approach
agents
- TCP
- MQTT

mindmap root)MarketAgent( Message Handler Receive Orderbook ::icon(fa fa-book) Receive Registration ::icon(fa fa-paper-plane) Event Obligation Inform Market opening ::icon(fa fa-door-open) Inform Market closing/result ::icon(fa fa-door-closed)

Market mechanisms

Different market mechanisms are modelled
- Pay-as-bid
- Pay-as-clear
- Pay-as-clear-complex
- Nodal-market (pyPSA)
- Redispatch (pyPSA)
- Pay-as-bid-with-contracts
- Two-step-clearing
- …
Modular design allows extensions

Concurrent markets and Continuous Auctions

Continuous auctions have an open orderbook
Continuity does not exist with computers
- Approximate through short clearing intervals
- Allows modeling by common bidding design

sequenceDiagram participant Market Market->>+Demand: Market opens for 1h Market->>+Generation: Market opens for 1h Generation-->>-Market: Bids [(vol: 20, cost: 300),(vol: 20, cost: 30)] Demand-->>+Market: Which bids available? Market->>-Demand: [(vol: 20, cost: 300),(vol: 20, cost: 30)] Demand-->>-Market: Bid (vol: 40, cost: 150) loop in discrete interval steps Market->>Demand: Market cleared Market->>Generation: Market cleared end Note right of Market: [(vol: 20, cost: 300), (vol: 20, cost: 150)] stay open positions

Eligible Lambda function

Some markets have additional constraints
- On bids
- On market participants
Configurable in market design through eligible_lambda function
- inplace function evaluated at registration
- only renewable actors can bid
- lambda unit: unit["technology"] in ["solar", "wind", ...] and unit["max_power"] >= 10

How to support OTC and custom contracts?

OTC is traded between parties
Virtual market containing bilateral trades
Contract execution is defined in market_mechanism
- Contract specifies executed code
- Trade gets accepted
- e.g. PPA - volume is calculated at the end

Second research question

In what way can different market mechanisms be represented in the simulation software?

Second research question

In what way can different market mechanisms be represented in the simulation software?

Third research question

How can key performance indicators of market mechanisms be used to evaluate specific market designs?

Key Performance Indicators

Grid optimization

Different scenarios
- Redispatch
- Nodal pricing
Works using pyPSA

Reinforcement Learning

Agents sample actions from dedicated learning role
Learning Role writes results into output database

Showcase

Market result analysis
Grid consideration
Key performance indicators
Dispatch of individual agent

Implementation

list of KPIs executed at end of simulation
configurable query building
Total Cost
- Name: total Cost
- Value: sum(price*demand_volume_energy)
- Table: market_meta
- Group bys: simulation

Third research question

How can key performance indicators of market mechanisms be used to evaluate specific market designs?

Third research question

How can key performance indicators of market mechanisms be used to evaluate specific market designs?

Summary of requirements

Functional

First research question
- Generic market design abstraction
Second research question
- Modelling of different market mechanisms
- Possible grid optimization through power flow calculation
Third research question
- Customizable output metrics
- Evaluation using key performance indicators

Summary of requirements

Functional

Generic market design abstraction
Modelling of different market mechanisms
Possible grid optimization through power flow calculation
Customizable output metrics
Evaluation using key performance indicators

Non-Functional

Availability/Open Source
Modularity
Interoperability with existing tools
Scalability

Non-functional requirements

Availability/Open Source
Modularity
Interoperability with existing tools
Scalability

Open-Source Software

Base for reproducibility and expandability

Open-Source Software

Concepts fully integrated into ASSUME. Two versions available:

independent PhD version

SPARKLE (Simulation Platform for Agent-based Research on energy marKet dynamics for Lower Emissions)

https://github.com/maurerle/sparkle

a maintained community version with other researchers and focus on RL

Modularity

Software should be modular so that it can be maintained easily and adapt to future features and requirements.

Modularity

Integration in ASSUME project
Customizable output metrics
Software patterns
- High cohesion
- Low coupling

mindmap root((modular simulation)) Market Clearing Pay as bid ::icon(fa fa-marker) Pay as clear ::icon(fa fa-clock) ... Bidding Strategies Naive MC strategy ::icon(fa fa-lightbulb) Learning strategy ::icon(fa fa-graduation-cap) Agent Units Powerplant ::icon(fa fa-sun) Storage ::icon(fa fa-battery-half) Demand ::icon(fa fa-plug) DSM ::icon(fa fa-building) Output metrics Total dispatch volume ::icon(fa fa-chart-line) Average price ::icon(fa fa-dollar-sign)

Component based structure

Strategy pattern for behavior

Builder pattern for scenario

State pattern for RL Buffer

Proxy for CSV/DB output

Interoperability

No standardized simulation input and output exists
Scenario in hierarchical data format (YAML, XML, JSON)
Time series data in tabular format (CSV, XLSX, DataFrame)

Interoperability

PyPSA based loader
- netCDF
- PyPOWER
- Pandapower
- PyPSA HDF5
AMIRIS [13]
- scenario/timeseries
- contracts
OEDS [14]
- ECMWF weather
- power plants from MASTR

Comparison study with AMIRIS at EEM2024

Know Your Tools - comparison study

Scalability

Runtime should profit from more cores
Distributed simulation
- Container-based
- Orchestration with docker swarm

Scalability / distributed clock

Process Intercommunication uses TCP
Main Clock orchestrates simulation

Scalability / performance evaluation

Cores	Speedup
1	1.00
2	1.82
4	3.18
8	5.21
16	7.32

Speedup = \frac{T_1}{T_n}

Scalability / non parallelizable market

Amdahl’s Law
- parallellizable part
- non-parallellizable part
Which amount limits scalability?

New design

Market abstraction found
Simulation approach
Iterative development

Model	Year	Market Design	OSS	Power Flow	Modularity	Interoperability	Scaled	Market Comparison
New Design	2023+	various	✓	✓ (pypsa)	✓	✓ (loaders)	✓ (mango)	✓

Limitations - what is not the focus

markets not covered by abstraction
optimized bidding behavior
most realistic parameterization for real simulation
investment planning/grid planning

Remaining TODOs

Better scalability interface
Improve output metrics
Usage scenarios
Finish thesis

Publication list

F. Maurer et al., “Know your tools - a comparison of open-source energy market simulation models” EEM2024
F. Maurer, J. Sejdija, V. Sander, “Decentralized Energy Data Storages through an Open Energy Database Server” NFDI4Energy Conference
F. Maurer et al., “Market Abstraction of Energy Markets and Policies – Application in an Agent-Based Modeling Toolbox“ Energy Informatics 2023
D. Stollenwerk, T. Franzke, F. Maurer et al., “Smarte Ladesäulen: Netz- und Marktdienliches öffentliches Laden“ Springer - Towards the New Normal in Mobility
F. Maurer, C. Rieke, R. Schemm, and D. Stollenwerk, “Analysis of an Urban Grid with High Photovoltaic and e-Mobility Penetration” Energies, vol. 16, no. 8, 8, p. 3380, Jan. 2023.
F. Maurer, “Framework To Provide A Simulative Comparison Of Different Energy Market Designs,” Energy Informatics, vol. 5, no. 2, p. 12, Sep. 2022.

Thank you for your attention!

Questions?

References

[1]

GridSingularity, “Grid Singularity Mission - Grid Singularity Wiki,” 2021. [Online]. Available: https://gridsingularity.github.io/gsy-e/documentation/. [Accessed: 15-Apr-2022].

[2]

M. Zade, S. D. Lumpp, P. Tzscheutschler, and U. Wagner, “Satisfying user preferences in community-based local energy markets — Auction-based clearing approaches,” Applied Energy, vol. 306, p. 118004, Jan. 2022.

[3]

J. Hiry, C. Kittl, F. Erlemeyer, and D. Schmid, “SIMONA – Agentenbasierte Netzsimulation für das intelligente Verteilnetz der Zukunft,” vol. Jahresbericht, pp. 40–41, Jan. 2018.

[4]

S. Glismann, “Ancillary Services Acquisition Model: Considering market interactions in policy design,” Renewable and Sustainable Energy Reviews, Apr. 2021.

[5]

J. Kochems and C. Schimeczek, “Agentenbasierte Modellierung von Lastmanagement im deutschen Stromsektor,” in Internationale Energiewirtschaftstagung (IEWT) 2021: Das Energiesystem nach Corona: Irreversible Strukturänderungen - Wie?, 2021.

[6]

USEF Design Team, “USEF - The Framework Explained,” p. 58, May 2021.

[7]

S. Battula, L. Tesfatsion, and T. E. McDermott, “An ERCOT test system for market design studies,” Applied Energy, vol. 275, p. 115182, Oct. 2020.

[8]

“Model PowerACE – REflex.” [Online]. Available: https://reflex-project.eu/model-coupling/powerace/. [Accessed: 10-Dec-2021].

[9]

M. Ketov, “MARKTSIMULATIONEN UNTER BERÜCKSICHTIGUNG DER STROM-WÄRME-SEKTORENKOPPLUNG,” p. 5, 2019.

[10]

T. Künzel and A. Weidlich, “Flexibility as an economic commodity in the intelligent energy system for the efficient integration of renewable energies,” 2015.

[11]

F. Maurer, K. K. Miskiw, R. R. Acosta, N. Harder, V. Sander, and S. Lehnhoff, “Market Abstraction of Energy Markets and Policies - Application in an Agent-Based Modeling Toolbox,” in Energy Informatics, vol. 14468, B. N. Jørgensen, L. C. P. Da Silva, and Z. Ma, Eds. Cham: Springer Nature Switzerland, 2023, pp. 139–157.

[12]

EPEX SPOT, “Trading Products | EPEX SPOT,” 2022. [Online]. Available: https://www.epexspot.com/en/tradingproducts. [Accessed: 09-May-2023].

[13]

F. Maurer, F. Nitsch, J. Kochems, C. Schimeczek, V. Sander, and S. Lehnhoff, “Know Your Tools - A Comparison of Two Open Agent-Based Energy Market Models,” in 2024 20th International Conference on the European Energy Market (EEM), 2024, pp. 1–8.

[14]

F. Maurer, J. Sejdija, and V. Sander, “Decentralized energy data storages through an Open Energy Database Server,” NFDI4Energy, Feb. 2024.

[15]

S. Loizidis, A. Kyprianou, and G. E. Georghiou, “Electricity market price forecasting using ELM and Bootstrap analysis: A case study of the German and Finnish Day-Ahead markets,” Applied Energy, vol. 363, p. 123058, Jun. 2024.