An agent-based decision support system for wholesale electricity market

doi:10.1016/j.dss.2007.05.007

Decision Support Systems

Volume 44, Issue 2, January 2008, Pages 425-446

https://doi.org/10.1016/j.dss.2007.05.007 Get rights and content

Abstract

Application software has been developed for analyzing and understanding a dynamic price change in the US wholesale power market. Traders can use the software as an effective decision-making tool by modeling and simulating a power market. The software uses different features of a decision support system by creating a framework for assessing new trading strategies in a competitive electricity trading environment. The practicality of the software is confirmed by comparing its estimation accuracy with those of other methods (e.g., neural network and genetic algorithm). The software has been applied to a data set regarding the California electricity crisis in order to examine whether the learning (convergence) speed of traders is different between the two periods (before and during the crisis). Such an application confirms the validity of the proposed software.

Introduction

Human decision-making is vital for the growth of an economy. Traders, who make various trading decisions, consist of an important component in a market. The bidding decision of a trader affects the profit of business. Among many different markets, the wholesale power market is a volatile market where a trader's bidding decision may affect the gain of a utility firm. The power industry is recently becoming competitive, unlike in the past where it was controlled by monopolistic utilities. A decentralized market environment is replacing the traditional centralized-operation approach. This business trend is called “deregulation of the electricity market.” [The importance of the electric power industry in research on decision support systems can be found in the special issue of this journal ([15] Vol. 40, Issues 3–4) organized by Oren and Jiang in 2005. The special issue contains 16 articles under the title of “Challenge of Restructuring the Power Industry” that have explored analytical aspects of the electric power industry.]

The deregulation allows new players to compete for providing wholesale electricity services by setting their own prices in an auction format, rather than negotiating with state regulators on a fixed price. Many wholesale power markets are directed towards liberalization and competition in the world. Along with the deregulation, many corporate leaders and policy makers face a difficulty in both predicting and understanding a price change of the wholesale electricity. The price change occurs due to many uncontrollable factors such as a change in weather condition, a demographic change and different trading strategies among traders. Software tools are needed for players in the power industry to predict the price change, to understand such market activities and to aid in their decision-making activities.

Many software systems have been developed for the purpose of aiding a trader in a wholesale electricity market. The shortcomings of existing systems are that they do not incorporate a transmission system in these algorithms and lack an estimation capability of the price fluctuation of electricity. To overcome the methodological difficulties, this study develops a decision support system (DSS), referred to as “MAIS (Multi Agent Intelligent Simulator)”, where software agents represent market entities such as generators, wholesalers, a market administrator, a network operator, and a policy regulator. The software agents have their own trading objectives and strategies. They can adjust their trading strategies in the simulation process based on previous trading efforts' success or failure. They also constantly observe current market price of electricity.

The proposed DSS provides a simulation-based numerical capability. That is the purpose of this study. This type of research is not present in the special issue (2005) organized by Oren and Jiang. [See Sueyoshi and Tadiparthi [28], [29] for a detailed description on the computer algorithm. Their research efforts are further extended in [30] that can investigate how a capacity limit on transmission influences the wholesale price of electricity.]

The structure of this article is organized as follows: The next section conducts a literature survey that indicates the position of the proposed DSS by comparing itself with other studies concerning on-line trading auctions. Section 3 describes both the architecture of MAIS and the software. Section 4 describes a market clearing algorithm that is incorporated into the proposed simulator. Section 5 documents the practicality of the proposed simulator, using a data set regarding the California electricity market. A concluding comment and future extensions are summarized in Section 6.

Section snippets

Previous works and existing software systems

Recently, Artificial Intelligence (AI) methods have been employed predominantly to solve various problems in decision making under uncertainty. Furthermore, many software systems have been developed to aid traders in their decision making in power trading. This section is subdivided into two parts. The first part surveys AI techniques used in the construction of DSS. The second part evaluates some of the software systems used for electricity trading.

Multi agent intelligent simulator

The MAIS consists of many software agents that interact with each other. They also interact with a power market (as an environment) by observing a price fluctuation of wholesale electricity.

Fig. 1 depicts the architecture of MAIS. There are five types of agents in MAIS: market administration, supply-side, demand-side, network operation, and utility policy making. The main objectives of an electricity market are to ensure the security of the power system, its efficient operation and further to

Market clearing algorithms

The market clearing process incorporated in the proposed simulator is structured by TSS. This type of auction process is used by PJM. Fig. 9 depicts the TSS at the t-th period that contains DA and RT along with these market components (fundamentals):

Structure of California electricity market

To document the practicality of the proposed MAIS, this study applies the simulator to a data set regarding the California electricity market from 1st April 1998 to 31st January 2001. The California electricity crisis occurred during the observed period. The market is divided into three zones for the purposes of pricing: NP15 is in the north, SP15 is in the south, and ZP26 is in the center of the state. The central zone (ZP26) has only 2 transmission links, one to Northern path (NP15) and one

Conclusion and future extension

MAIS can be used for analyzing and understanding a dynamic price change in the US wholesale power market. Traders can use the software as an effective DSS tool by modeling and simulating a power market. The software uses various features of DSS by creating a framework for assessing new trading strategies in a competitive electricity trading environment. The practicality of the software is confirmed by comparing its estimation accuracy with those of other methods (e.g., neural network and

Acknowledgement

This research is supported by Telecommunication Advancement Foundation.

Toshiyuki Sueyoshi is a full professor for Department of Management at New Mexico Institute of Mining and Technology in USA. He is also a visiting full professor for Department of Industrial and Information Management at National Cheng Kung University in Taiwan. He obtained his Ph.D. from University of Texas at Austin. He has published more than 150 articles in well-known international journals.

References (37)

B.S. Ahn et al.
Integrated methodology of rough set theory and artificial neural network for business failure prediction
Expert Systems with Applications
(2000)
M. Beynon et al.
A new paradigm for computer-based decision support
Decision Support Systems
(2002)
T. Bui et al.
Agent-based framework for building decision support systems
Decision Support Systems
(1999)
C. Changchit et al.
Supporting managers' internal control evaluations: an expert system and experimental results
Decision Support Systems
(2001)
R.R. Hashemi et al.
A hybrid intelligent system for predicting bank holding structures
Journal of Operational Research
(1998)
T.P. Liang et al.
Framework for applying intelligent agents to support electronic trading
Decision Support Systems
(2000)
T.E. McKee et al.
Genetic programming and rough sets: a hybrid approach to bankruptcy classification
European Journal of Operational Research
(2002)
M. Ozbayrak et al.
A knowledge-based decision support system for the management of parts and tools in FMS
Decision Support Systems
(2003)
S. Park et al.
Designing multi-agent systems: a framework and application
Expert Systems with Applications
(2005)
S. Schocken et al.
Neural networks for decision support: problems and opportunities
Decision Support Systems
(1994)

N.H. Sung et al.

Knowledge assisted dynamic pricing for large-scale retailers

Decision Support Systems

(2000)

Y.F. Wang

Mining stock price using fuzzy rough set system

Expert Systems with Applications

(2003)

R.L. Wilson et al.

Bankruptcy prediction using neural networks

Decision Support Systems

(1994)

M.E. Yahia et al.

Rough neural expert systems

Expert Systems with Applications

(2000)

J. Zeleznikow et al.

Using soft computing to build real world intelligent decision support systems in uncertain domains

Decision Support Systems

(2001)

R. Zimmerman et al.

A web-based platform for experimental investigation of electric power auctions

Decision Support Systems

(1999)

I. Acronymics

Agentbuilder 1.4

(2004)

K.J. Adams et al.

Knowledge discovery from decision tables by the use of multiple-valued logic

Artificial Intelligence Review

(2003)

Cited by (61)

Machine learning in marketing: A literature review, conceptual framework, and research agenda
2022, Journal of Business Research
In recent years, machine learning (ML) and artificial intelligence (AI) have attracted considerable attention in different industry sectors, including marketing. ML and AI hold great promise for making marketing intelligent and efficient. In this study, we conduct a literature review of academic journal studies on ML in marketing applications and propose a conceptual framework highlighting the main ML tools and technologies that serve as the foundation of ML applications in marketing. We use the 7Ps marketing mix, that is, product, price, promotion, place, people, process, and physical evidence, to analyze these applications from 140 selected articles. The applications are supported by various ML tools (text, voice, image, and video analytics) and techniques such as supervised, unsupervised, and reinforcement learning algorithms. We propose a two-layer conceptual framework for ML applications in marketing development. This framework can serve future research and provide an illustration of the development of ML applications in marketing.
Harnessing business intelligence in smart grids: A case of the electricity market
2018, Computers in Industry
This paper discusses analytical aspects of smart grids and offers insights into the development of a business intelligence solution for the electricity market. The goal is to design a system that provides an emerging electricity market with the necessary data flows and information for forecasting, data analysis and decision making, leading to better business results and more control over the market. By employing a methodology specifically suited to the electricity market domain, we designed a business intelligence solution for the Serbian electricity market operator “Elektromreža Srbije”. The research results show that the proposed approach leads to more effective market management in data-rich smart grid environments, while still being dynamic enough to adapt to frequent rule changes in the still developing grids and their markets.
Measuring scale efficiency and returns to scale on large commercial rooftop photovoltaic systems in California
2017, Energy Economics
Electricity generated by solar Photovoltaic (PV) power systems is emerging as one of the most promising cleaner alternatives to replace conventional energy sources, such as coal, oil, and natural gas. This study examines managerial sources of operational efficiency or inefficiency on 855 large commercial rooftop PV power systems in California by examining both scale efficiency and Returns to Scale (RTS). For the research purpose, this study utilizes Data Envelopment Analysis (DEA) as a methodology to assess the scale measures. A difficulty in applying DEA to the performance analysis of PV power systems is that it contains uncontrollable variables such as ambient temperature and solar irradiation, all of which are influenced by weather. It is also widely known that PV power systems cannot produce electricity during night, so being unable to serve as a base load. Thus, these uncontrollable factors often influence the performance of PV power systems. Paying attention to the effects of those uncontrollable factors, this study discusses how to measure scale efficiency and RTS within the framework of DEA. Then, as an application of the proposed approach, this study examines the performance of large commercial rooftop PV power systems in California. This study finds that they belong to increasing (48), constant (807) and decreasing (0) RTS, implying that most of the PV power systems in California have been operating in their appropriate sizes. This further indicates that the operational inefficiency of PV power systems is due to managerial inappropriateness, not their sizes for generation. Thus, there is a space to improve operational efficiency of rooftop PV systems under constant RTS, through managerial efforts such as increase in transmission capacity and operational flexibility to improve solar penetration, and routine maintenance to minimize detrimental effects of dust deposition.
Measurement of returns to scale on large photovoltaic power stations in the United States and Germany
2017, Energy Economics
This study discusses how to classify the type of RTS (Returns to Scale) in the framework of DEA production analysis under the assumption on a unique optimal solution on projection and a reference set. The RTS measurement is classified into two categories: input-based RTS and output-based RTS. The two measures depend upon the sign of an intercept (σ) of a supporting hyperplane on a production possibility set. In the DEA framework, multiple solutions usually occur on the intercept. This study discusses how to handle such an occurrence of multiple intercepts within the framework of the input-based and output-based RTS classifications. As an application, this study applies the proposed approach to classify the type of RTS on large photovoltaic power stations, often referred to as “mega solar parks”, in the United States (US) and Germany. The input-based RTS measurement identifies that US photovoltaic power stations are classified into increasing (64 stations: 80.0%), constant (5 stations: 6.3%) and decreasing (11 stations: 13.8%), respectively. German photovoltaic power stations are classified into increasing (73 stations: 91.3%), constant (7 stations: 8.8%) and decreasing (0 station: 0.0%), respectively. The finding is confirmed by the output-based RTS measurement on the German photovoltaic power stations. However, the output-based RTS on US photovoltaic power stations shows an opposite result. The US photovoltaic power stations are classified into increasing (26 stations: 32.5%), constant (6 stations: 7.5%) and decreasing (48 stations: 60.0%), respectively. The decreasing RTS is because the US photovoltaic power stations have less efficiently operated than those of Germany in such a manner that they cannot produce sufficient outputs from a given level of inputs in such a manner that they have failed to utilize a scale merit for their operations.
Multi-step prediction of experienced travel times using agent-based modeling
2016, Transportation Research Part C: Emerging Technologies
This paper develops an agent-based modeling approach to predict multi-step ahead experienced travel times using real-time and historical spatiotemporal traffic data. At the microscopic level, each agent represents an expert in a decision-making system. Each expert predicts the travel time for each time interval according to experiences from a historical dataset. A set of agent interactions is developed to preserve agents that correspond to traffic patterns similar to the real-time measurements and replace invalid agents or agents associated with negligible weights with new agents. Consequently, the aggregation of each agent’s recommendation (predicted travel time with associated weight) provides a macroscopic level of output, namely the predicted travel time distribution. Probe vehicle data from a 95-mile freeway stretch along I-64 and I-264 are used to test different predictors. The results show that the agent-based modeling approach produces the least prediction error compared to other state-of-the-practice and state-of-the-art methods (instantaneous travel time, historical average and k-nearest neighbor), and maintains less than a 9% prediction error for trip departures up to 60 min into the future for a two-hour trip. Moreover, the confidence boundaries of the predicted travel times demonstrate that the proposed approach also provides high accuracy in predicting travel time confidence intervals. Finally, the proposed approach does not require offline training thus making it easily transferable to other locations and the fast algorithm computation allows the proposed approach to be implemented in real-time applications in Traffic Management Centers.
Energy Internet: The business perspective
2016, Applied Energy
Energy Internet is a new development form of energy system. It realizes the integration of energy flow, information flow and business flow. More and more business model and service model innovations are stimulated in Energy Internet. In this paper, we present a systemic study of Energy Internet from the business perspective. We first propose the evolution stages of energy systems. Since they were invented in the second industrial revolution period, energy systems have mainly experienced four stages, i.e., decentralized energy system, centralized energy system, distributed energy system and smart & connected energy system. Energy Internet is the innovative representation of energy systems in the fourth development stage. We also introduce some key concepts in Energy Internet, including prosumer, microgrid, Virtual Power Plant (VPP), smart grid and smart energy. Then the business values of Energy Internet are discussed from the energy big data analytics perspective. Finally, the business research agendas of Energy Internet are pointed out from five aspects, i.e., strategic issues, data issues, behavioral issues, security issues and regulatory issues.

View all citing articles on Scopus

Gopalakrishna Reddy Tadiparthi received a B.E. degree in Computer Science from University of Madras, India, in 1999 and worked as a Network Engineer at Satyam Infoway Limited (SIFY), India till 2002. He received a M.S. degree in Computer Science from New Mexico Institute of Mining and Technology in 2003 and is currently a Ph.D. candidate in the Department of Computer Science.

View full text

An agent-based decision support system for wholesale electricity market

Abstract

Introduction

Section snippets

Previous works and existing software systems

Multi agent intelligent simulator

Market clearing algorithms

Structure of California electricity market

Conclusion and future extension

Acknowledgement

Expert Systems with Applications

Decision Support Systems

Decision Support Systems

Decision Support Systems

Journal of Operational Research

Decision Support Systems

European Journal of Operational Research

Decision Support Systems

Expert Systems with Applications

Decision Support Systems

Decision Support Systems

Expert Systems with Applications

Decision Support Systems

Expert Systems with Applications

Decision Support Systems

Decision Support Systems

Agentbuilder 1.4

Knowledge discovery from decision tables by the use of multiple-valued logic

Artificial Intelligence Review