Research Article  Open Access
Research on Power Producer’s Bidding Behavior Based on the BestResponse Dynamic Model
Abstract
As China’s electricity market is facing many problems, the research on power producer’s bidding behavior can promote the healthy and sustainable development of China’s electricity market. As a special commodity, the “electricity” possesses complicated production process. The instable market constraint condition, nonsymmetric information, and a lot of random factors make the producer’s bidding process more complex. Bestresponse dynamic is one of the classic dynamic mechanisms of the evolutionary game theory, which applies well in the repeated game and strategy evolution that happen among a few bounded rational players with a quick learning capability. The bestresponse dynamic mechanism is employed to study the power producer’s bidding behavior in this paper, the producer’s bestresponse dynamic model is constructed, and how the producers would engage in bidding is analyzed in detail. Taking two generating units in South China regional electricity market as the example, the producer’s bidding behavior by following the producer’s bestresponse dynamic model is verified. The relationships between the evolutionarily stable strategy (ESS) of power producer’s bidding and the market demand, and ceiling and floor price as well as biding frequency are discussed in detail.
1. Introduction
In the 1970s, when ecologists Maynard Smith and Price studied the ecology evolution phenomenon, they combined the biological evolutionism with the game theory and then came up the theory of evolutionarily stable strategy (ESS) [1, 2], which marked the birth of evolutionary game theory. Compared with the traditional game theory, the evolutionary game theory improves the Perfect Rationality assumption and combines the traditional game theory with the dynamic evolution process to describe how the evolutionary relationship of gaming would develop over time, which overcomes the perfect rationality paradox in both neoclassical economics and game theory. Thus, it provides a new analysis method for economic research [3]. In the middle of the 1990s, Fudenberg and Levine elaborated in detail the specific content of game learning theory based on the ideas of “learning” and “evolution” [4]. This theory was accepted by many economists in a very short time and achieved remarkable accomplishments in economic analysis [5, 6]. From the 1980s, the evolutionary game theory has been widely used in economic field, such as the change of macroeconomic system [7], enterprise competition [8], and so on.
When it comes to group decisionmaking, the player often observes others firstly and then learns from the game history to adjust his game strategy. The key point of evolutionary game theory is to identify the strategy adjustment path. Currently, there are three types of evolutionary decisionmaking mechanisms: the mechanism based on the bestresponse dynamic and replicate dynamic model, the mechanism based on random process or swarm intelligence optimization algorithm, and the mechanism based on neural network and reinforcement learning. Among those, the bestresponse dynamic model and the replicate dynamic model based on biological evolution are the most commonly used dynamic decisionmaking mechanisms [9–11].
Since the implementation of the Electric Power System Reformation Plan approved by the China State Council in March 2002, the electricity market reform has gained primary achievement. One outcome is that China has fully separated the power generation from the power transmission. Nowadays, the power grids in China are mainly operated by the State Grid Corporation of China (SGCC) and the China Southern Power Grid Corporation (CSPGC). The auxiliary power companies include China Power Engineering Consulting Group, China Hydropower Engineering Consulting Group, China Water Conservancy, Hydropower Construction Group, and China Gezhouba Corporation. This indicates that the separation of the assistant industry from the main power industry has been achieved to some extent [12, 13].
Although China’s electricity market reform has made some achievements, there still remain many problems. China’s top five largescale power generation groups and some independent power generation companies with considerable size have initially formed the oligopolistic competitive pattern on the power generation side, but the regional electricity markets located in North China, Northeast China, Northwest China, East China, and Central China are just established and currently do not have fullfledged trading rules or settlement mechanisms. Besides, some serious problems have emerged at some regional markets, and even some ongoing pilot projects have been interrupted. To pursue greater profits, all the power generation companies have motives to increase the electricity price at the regional electricity market, which will naturally lead to the strategic bidding of many power generation companies. This will bring harm to the safe operation of power grid as well as the price stability [14, 15]. To solve those problems caused by marketoriented reform, China not only needs to prefect the trading rules and establish reasonable trading mechanisms but also needs to study the producer’s bidding behavior.
Because of the specificity of electricity as a commodity and the complexity of its production, the traditional game model cannot be applied well in the electricity market. Meanwhile, the instable market constraints, information asymmetry, and plenty of random factors affect the biding process, which make this issue more complex. Therefore, in this paper, according to the actual situation of China’s electricity market, the bestresponse dynamic model of oligopolistic power producer’s bidding is constructed based on the assumption that all producers have bounded rationality, and then the relationships between the evolutionarily stable strategy (ESS) of power producer’s bidding and the market demand and ceiling and floor price as well as biding frequency are discussed.
The paper is organized as follows: Section 2 establishes the bestresponse dynamic model of power producer’s bidding; Section 3 performs the producer’s bidding behavior analysis; taking two biding generating units in South China regional electricity market as an example, the empirical analysis is performed in Section 4; Section 5 concludes this paper.
2. BestResponse Dynamic Model of Power Producer’s Bidding
2.1. Basic Theory of BestResponse Dynamic Model
The bestresponse dynamics is usually applied in the adjustment process of a repeated game among players who have rapid learning ability and bonded rationality. The rapid learning ability means that the players can make accurate postevaluation on the results of different strategies and then adjust their strategies accordingly although their abilities of judgment and foresight are a bit poor under the complicated situation. Therefore, when the former gaming result is given, each player can identify the bestresponse strategy compared with the former strategies adopted by other players.
According to the theory of Fudenberg and Levine, the bestresponse dynamic equation of player can be represented as [16] where is the probability distribution of strategy adopted by player at time ; is the best response of player against his competitors’ strategies; and is the ratio of the players who adopt the bestresponse strategy in all the players at time . The rest of the players continue to choose their strategies at time .
Equation (1) is the general mathematical description of adjustment process of repeated game between players who have rapid learning ability and bonded rationality.
2.2. The Mechanism and Assumption of BestResponse Dynamic Model of Power Producer’s Bidding
The random bestresponse dynamic model of producer’s bidding adjustment was proposed by Larson and Salant, the mechanism of which is that each producer tends to adjust his bidding strategy according to the competitors’ previous bidding strategies [17].
Suppose that represents the probability distribution of producer ’s bidding at time . The probability distribution of producer ’s bidding is the arithmetic weighted mean which is calculated by and . is the probability distribution of producer ’s bidding at time and is the correction term. represents the bestresponse made by producer when it comes to competitors. Therefore, producer would adjust his bidding price according to where is the ratio of the players choosing the bestresponse strategy, and it also represents the weight of best response. The more the profit the competitor gets in prior time, the bigger the would become. After a round of bidding, producers could obtain the probability distribution of competitor’s bidding price at this round, and then they compute the corresponding best response based on this knowledge.
Some assumptions need to be made when performing the bestresponse dynamic model of power producer’s bidding. Suppose that there are producers competing in one regional electricity market, and the maximum and minimum market trading prices are and , respectively. At bidding time , the market demand is ; the producer’s bidding prices are ordered, and the last producer’s bidding price that meets the market demand is named as the market clearing price .
The producer adjusts the bidding price by learning the previous bidding information. There are two main factors that affect producer ’s bidding price at time , which are as follows.(1)Generating Cost . is a variable related to the bidding power volume (power generation) , and suppose that where represents the producer’s capacity cost and represents the change rate of electricity cost.(2)Competitor’s Bidding Price . At each round of bidding, producer does not know the competitor’s bidding price , but he can estimate that the competitor’s bidding price obeys the probability distribution function which has density function on interval .
2.3. Producer’s Profit Function
If a producer’s bidding price is lower than or equal to the clearing price, he can sell out all the declaratory power generation at the unified clearing price. However, if the producer’s bidding price is higher than the clearing price, his power generation will not be sold out, and then his profit will be equal to zero. The profit function of producer at time can be represented as Then, the expected profit of producer at time can be represented as where is the probability of producer ’s successful bidding; is the possibility of bidding failure. To calculate and , and are supposed as the possibility of producer ’s bidding price higher and lower than his competitor ’s bidding price, respectively.
According to [18], if producer has a bidding failure and producers all have successful biddings, then there are at least producers whose bidding prices are lower than producer ’s bidding price . Hence, the possibility of producer ’s failure bidding and the possibility of successful bidding can be determined as follows: So, the expected profit of producer at time can be calculated by
2.4. Dynamic Adjustment Process of Producer’s Best Response
After one round of bidding, the producer can learn some information about the competitors’ latest bidding price and then adjust his own bidding price, so that he can make the best response against his competitors. Suppose that represents the bidding price adjustment of producer at time . The adjustment mechanism is as follows: firstly, the producer calculates his profit according to the previous bidding price ; secondly, the producer adjusts the previous bidding price under the scenario that other producers will not change their bidding strategies and then calculates the corresponding profit (this profit is not the actual profit); finally, is compared with . If , the producer can gain more profit by the bidding price adjustment, so the producer will make the bidding price adjustment; namely, ; otherwise, the producer will not make the bidding price adjustment; namely, . Therefore, the bidding price adjustment process of producer can be represented as
Figure 1 shows the adjustment process of producer ’s price bidding strategy, which is also called the learning process of producer ’s price bidding. Once the producer determines his initial price, the expected profit can be calculated, and the bidding price adjustment can be made based on the bestresponse dynamic model at each price bidding round.
2.5. Judgment Criteria of Evolutionarily Stable Strategy for Producer’s Bidding Behavior
Evolutionarily stable strategy (ESS) is an important concept in the game learning theory, which reflects the achieved equilibrium state after the bestresponse dynamic adjustment process. According to the connotation of ESS, the ESS of bestresponse dynamic adjustment on producer’s bidding price is as follows.
Suppose is the bidding price strategy of producers, is the ratio of the producers who adopt the bidding strategy , and then is the ratio of the producers who adopt price strategy .
If the producer’s profit satisfies , then the bidding price strategy is defined as one of the ESS during the bestresponse dynamic adjustment process. represents the distribution of producers adopting strategy and .
An evolutionarily stable strategy should meet the following requirements.(1)The proportion of individuals adopting this strategy keeps constant, which means the value of is constant.(2)This stable state must have robustness against the slight disturbance, which means the system can automatically recover to the evolutionarily stable state from the unstable state.
Therefore, the ESS of bestresponse dynamic adjustment on producer’s bidding price should satisfy the following two conditions.(1)The profit of producer keeps the same no matter whether the producer adjusts the bidding price or not; namely, (2)Even though there exists a slight bidding strategy disturbance which makes the bidding price deviated from the stable state , the bidding price can still go back to the stable state after the bestresponse dynamic adjustment; namely,
3. Producer’s Bidding Behavior Analysis
The bestresponse dynamic model applies well in the gaming behavior that involves a few players who have strong learning ability. Meanwhile, the producers in the oligopolistic electricity power have the characteristics of small number, large scale, and strong information searchinganalyzingprocessing capability. By learning the historical market information and predicting the development trend, the producer can estimate both the competitors’ bidding prices and profits and then acts properly against the competitors’ bidding strategies. Therefore, the bestresponse dynamic model of power producer’s bidding can be used to study the bidding behavior of oligopolistic producer in the electricity market and the price bidding trend.
Suppose that there are two exact oligopolistic producers in one regional electricity market. Due to the symmetry, this paper only needs to study one oligopolistic producer i, and the other oligopolistic producer is the competitor of producer . These two producers have the same production capability and cost . Set the ceiling price and the floor price in the electricity trading market as and , respectively, and the bidding price obeys the uniform distribution in interval ; namely, . The market demand is .
3.1. BestResponse Dynamic of Two Oligopolistic Producers
According to the above suppositions and the bestresponse dynamic model, the profit of producer can be represented as When , the producer will continue to make the adjustment on bidding price.
According to (9)(10), we can get Then,
Equation (13) is the mathematical expression of ESS of producer ’s bidding through the bestresponse dynamic adjustment.
3.2. The Relationship between ESS and Market Demand, the Ceiling Price, and the Floor Price
Suppose that is a constant, which means the bidding adjustment does not have an effect on ESS. When the ceiling price, floor price, and generating cost are given, the relationship between ESS of producer ’s bidding and market demand is determined, which is shown in Figure 2.
In Figure 2, the solid line shows the relationship between and when RMB/MWh, RMB/MWh, RMB/MWh, and . When the ceiling price () increases to 300 RMB/MWh, the relationship between and is shown as the dashed line.
Just as shown in Figure 2, when increases, goes up; when increases, goes up. This indicates that the producer’s bidding price will become higher with the increase of electricity demand in the regional electricity market. Hence, controlling the ceiling price is an effective method to keep the oligopolistic producer’s bidding price under limits. When the ceiling price is set at low level, the producer’s bidding price can be controlled, while the excessive celling price cannot play a role in controlling the producer’s price bidding behavior.
Meanwhile, the floor price can also affect the producer’s bidding strategy, but its effect is quite special: when the market demand is small (), the higher the floor price is, the higher the producer’s bidding price will become; yet when the market demand is large (), the higher the floor price is, the lower the producer’s bidding price will become. This indicates that when the electricity demand is small, the floor price can play a role in avoiding the virulent price bidding behavior of producers; but when the electricity demand is large, the floor price will not work.
3.3. The Relationship between ESS and Bidding Frequency
Suppose that the initial bidding price of producer is and the bidding frequency is . Then, the ESS can be represented as where , which can be derived from (13).
Then, we can get
When the marker demand, ceiling price, floor price, and producer’s initial bidding price are given, the relationship between bidding frequency and ESS can be discussed. Suppose , RMB/MWh, RMB/MWh, RMB/MWh, and RMB/MWh; then we can get
The relationship between ESS of producer’s bidding and bidding frequency is shown in Figure 3. As shown in Figure 3, the ESS of producer’s bidding will go up with the increase of bidding frequency and eventually converges to bidding price strategy with the maximum profit. In the regional electricity market, adding the number of market trading will increase the bidding frequency, and the producer can gradually adjust his bidding price by learning market information and analyzing competitor’s bidding strategy until the maximum profit can be obtained.
Moreover, RMB/MWh. This bidding price is a Nash equilibrium price of producer’s bidding by using the traditional game theory, which indicates that, through the longterm and multiple bidding price adjustment, the bounded rationality producer who possesses limited information can find the optimal biding price with the maximum profit. However, in this bidding price adjustment process, every bidding price offered by the producer may not be the optimal one, which in return backs up the ESS connotation in the game learning theory; namely, the game equilibrium is the result of longterm seeking optimization of boundedrationality participants.
4. Empirical Analysis
4.1. Sample Data
Two competitive generating units from South China regional electricity market are selected. Considering the limitation of essential data, the sample range includes 24 time points of relevant indicators of generating units. The cost and production capacity of these two competitive generating units are listed in Table 1, the declaratory electricity power and limited prices are listed in Table 2, and the declaratory electricity prices of generating units are listed in Table 3.



4.2. Bidding Behavior Analysis
4.2.1. The Relationship between ESS and the Market Demand, the Ceiling Price, and the Floor Price
In the region electricity market, the electricity demand fluctuates over time, which is shown as the increase or decrease in the declaratory electricity power at different periods of time. Just as shown in Figure 4, when the electricity demand increases, the producer’s bidding price tends to go up, while the producer tends to go for the low bidding price when the electricity demand is relatively low.
When the ceiling price is set at a low level, the producers will offer a low price, which makes the overall bidding price become low; but when the ceiling price is set at a high level, the producer’s optimal bidding price will rise. So, the ceiling price could not inhibit the motive of producer to raise the electricity price, which is shown in Figure 5.
The floor price will also affect the producer’s bidding strategy. Due to the fact that the sample data is selected from the regional electricity market, its market demand is relatively small compared with the overall market demand. So, the multiple of market demand to producer’s production capability should be between 0 and 1. Under this situation, the higher the floor price is, the higher the producer’s bidding price is, and the adjustment of producer’s bidding price will be consistent with the adjustment of the floor price, which is shown in Figure 6.
4.2.2. The Relationship between ESS and Bidding Frequency
According to the above analysis, the ESS of producer’s bidding will go up with the increase of the bidding frequency and eventually converges into the bidding price strategy expressed as that possesses the maximum profit. The calculation result of the relationship between ESS and bidding frequency based on the sample data is shown in Figure 7. From Figure 7, we can see that the producer’s bidding price converges to the most profitable bidding strategy after a fourtime adjustment.
5. Conclusions
The producer’s bidding strategy based on the bestresponse dynamic mechanism is studied and the bestresponse dynamic model of producer’s bidding behavior is constructed in this paper. Taking two generating units in South China regional electricity market as the example, the monopolistic producer’s bidding behaviors are empirically studied, and some conclusions are drawn as follows.(1)With the increase of electricity power demand, the oligopolistic producer tends to raise his bidding price. If the bidding behavior cannot be restrained, when the market demand goes near the producer’s supply capacities, all producers will raise the bidding price to a very high level. This conclusion has been proved by the power crisis in California.(2)The ceiling price has some certain effects on inhibiting the motive of producers to raise the electricity price, and the producer’s overall bidding price will go with the ceiling price. This implies that setting reasonable celling price should not only consider how to keep the producers away from raising the bidding price, but also consider how to motive the producers in terms of profit.(3)When the electricity power demand is small, the floor price can play a role in avoiding the virulent price bidding behavior of producers. But when the electricity demand becomes large, the floor price will not work.(4)When the number of market trading increases, the producer’s bidding frequency will increase, and the producer can gradually adjust his bidding price by learning market information and analyzing competitors’ bidding strategies until the maximum profit can be obtained.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This study is supported by the National Natural Science Foundation of China (Project no. 71373076) and the Humanities and Social Science project of the Ministry of Education of China (Project no.11YJA790217). The authors are grateful to the editor and two anonymous reviewers for their suggestions in improving the quality of the paper.
References
 J. M. Smith and G. R. Price, “The logic of animal conflict,” Nature, vol. 246, no. 5427, pp. 15–18, 1973. View at: Publisher Site  Google Scholar
 J. M. Smith, Evolution and Theory of Games, Cambridge University Press, New York, NY, USA, 1982.
 M. A. Nowak, Evolutionary Dynamics, Belknap Press, Cambridge, Mass, USA, 2006.
 D. Fudenberg and D. K. Levine, The Theory of Learning in Games, The MIT Press, Cambridge, Mass, USA, 2000.
 X.M. Chen, W.D. Meng, and D.J. Hu, “Evolutionary game analysis on opportunistic behavior in international joint venture,” System Engineering Theory and Practice, vol. 29, no. 2, pp. 53–62, 2009. View at: Google Scholar
 Y. Xu, B. Hu, and R. Qian, “Analysis on stability of strategic alliances based on stochastic evolutionary game including simulations,” System Engineering Theory and Practice, vol. 31, no. 5, pp. 920–926, 2011. View at: Google Scholar
 J. Miekisz, “Evolutionary game theory and population dynamics,” in Multiscale Problems in the Life Sciences, vol. 1940 of Lecture Notes in Mathematics, pp. 269–316, 2008. View at: Google Scholar
 C. Homburg, A. Fürst, T. Ehrmann, and E. Scheinker, “Incumbents' defense strategies: a comparison of deterrence and shakeout strategy based on evolutionary game theory,” Journal of the Academy of Marketing Science, vol. 41, no. 2, pp. 185–205, 2013. View at: Publisher Site  Google Scholar
 D. Foster and P. Young, “Stochastic evolutionary game dynamics,” Theoretical Population Biology, vol. 38, no. 2, pp. 219–232, 1990. View at: Google Scholar
 W.B. Liu and X.J. Wang, “Multiagent reinforcement learningmodel in evolutionary games,” System Engineering Theory and Practice, vol. 29, no. 3, pp. 28–33, 2009. View at: Google Scholar
 D. Fudenberg and J. Tirole, Game Theory, The MIT Press, Cambridge, Mass, USA, 1991.
 Q. Wang and X. Chen, “China's electricity marketoriented reform: from an absolute to a relative monopoly,” Energy Policy, vol. 51, pp. 143–148, 2012. View at: Publisher Site  Google Scholar
 J. A. Cherni and J. Kentish, “Renewable energy policy and electricity market reforms in China,” Energy Policy, vol. 35, no. 7, pp. 3616–3629, 2007. View at: Publisher Site  Google Scholar
 X. Zhang, X. Wang, J. Wang, and Z. Hu, “Research on the producer's longterm power distribution strategy,” Proceedings of the CSEE, vol. 25, pp. 6–12, 2005. View at: Google Scholar
 C. Peng, H. Sun, J. Guo, and G. Liu, “Multiobjective optimal strategy for generating and bidding in the power market,” Energy Conversion and Management, vol. 57, pp. 13–22, 2012. View at: Publisher Site  Google Scholar
 D. Fudenberg and D. K. Levin, Game Learning Theory, China Renmin University Press, Beijing, China, 2004.
 N. Larson and D. Salant, Equilibrium in Wholesale Electricity Markets, Energy Research Center of the Netherlands (ECN), 2003.
 Y. Ren, X. Zou, and X. Zhang, “Generating company’s bidding game model based on the toplevel sealed auction,” Journal of Systems Engineering, vol. 18, pp. 248–254, 2003. View at: Google Scholar
Copyright
Copyright © 2014 Jingqi Sun et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.