بررسی انتقادی از مدل های بازار عمده فروشی برق مبتنی بر عامل

The complexity of electricity markets calls for rich and flexible modeling techniques that help to understand market dynamics and to derive advice for the design of appropriate regulatory frameworks. Agent-Based Computational Economics (ACE) is a fairly young research paradigm that offers methods for realistic electricity market modeling. A growing number of researchers have developed agent-based models for simulating electricity markets. The diversity of approaches makes it difficult to overview the field of ACE electricity research; this literature survey should guide the way through and describe the state-of-the-art of this research area. In a conclusive summary, shortcomings of existing approaches and open issues that should be addressed by ACE electricity researchers are critically discussed.

Because of the technical nature of the traded good, electricity markets rank among the most complex of all markets operated at present. Supply and demand have to be balanced in real time, considering transmission limits and unit commitment constraints. The electricity sector is characterized by multiple interlinked markets: fuel markets, markets for day-ahead scheduling and those for real-time dispatch or balancing energy, bilateral trading and auxiliary markets e.g. for emission allowances. Many energy firms are vertically integrated and act on several markets simultaneously, thus further complicating their trading strategies. Besides, and given the oligopolistic structure of almost all electricity markets, participants have the potential to exert market power in many of these markets. These complexities drive most classical modeling methods to their limits. Equilibrium models either do not consider strategic bidding behavior or assume that players have all relevant information about the other players' characteristics and behavior; they also disregard the consequences of learning effects from daily repeated interaction (Rothkopf, 1999). Game theoretical analysis is usually limited to stylized trading situations among few actors, and places rigid – oftentimes unrealistic – assumptions on the players’ behavior. Human-subject experiments can be applied to electricity market research only with difficulties, because some expertise is necessary to realistically imitate the bidding behavior of a power generator. Thus, for many questions relevant in electricity market research, human-subject experiments are not an appropriate method. Given the complexity of the electricity sector and also its high importance for a competitive economy, researchers and practitioners are increasingly willing to try new modeling methods in order to gain insights into various aspects of power markets. Agent-based (AB) modeling is one appealing new methodology that has the potential to overcome some shortcomings of traditional methods. Within the last ten years, more and more researchers have been developing electricity market models with adaptive software agents. This field of research is still growing and maturing. Some first attempts have already been revoked, others have gained popularity. As the number of AB electricity models that have been published in journals starts to increase, and given the high attractiveness of agent-based approaches among researchers, this survey might help to get a sense of the state-of-the-art of the whole research field.1 This literature review is supposed to guide newcomers or interested researchers through the intricate research field and points out the weaknesses and open issues that current approaches face. It is structured as follows: Section 2 gives a brief introduction to the methodology of Agent-Based Computational Economics (ACE); Section 3 presents the approaches and findings of relevant scientific papers in ACE electricity market research, and Section 4 summarizes the contributions made by the papers, points out some shortcomings of the current state-of-the-art, and suggests some lines of future work in the research field. Finally, Section 5 concludes.

This paper has critically reviewed a considerable amount of relevant papers in agent-based electricity market research. Table 1 summarizes the core characteristics of the cited work and displays the similarities and differences between the approaches. In Section 4 we have identified some of the current problems facing this research methodology that require further effort and a consolidation of the approaches pursued by different research groups. Especially sound argumentations for the choice of specific learning algorithms, more careful and well documented validation and verification procedures as well as the appropriate publication of details of concrete simulation models are crucial for the further development of agent-based electricity market modeling. Despite the open issues and problems, AB electricity research has been successful in recent time. Many AB researchers have successfully replicated core characteristics of today's electricity markets using models with adaptive, self-seeking agents. With a decrease of heterogeneity between competing models, and with increasing consensus on important methodological questions, the field of AB modeling can soon become one major strand of research for the analysis of complex electricity systems.