基于预测惩罚和联盟嫉妒的分布式能源合作博弈模型

Cooperative game model in distributed power plants based on prediction penalty and alliance envy

本研究提出一种基于分布式能源电厂的盈缺电量与运输成本进行联盟的合作博弈模型,探讨分布式能源在合作下的效用最大化问题。通过该模型减少因分布式能源预测发电量与实际发电量差值问题产生的惩罚,同时考虑联盟间的嫉妒因素精确每个分布式能源电厂效用,最后依据Shapley值法为各分布式电厂分配效用。数值结果表明,发电厂之间的合作不仅可以降低分布式能源发电量不确定性带来的惩罚,同时可以减少联盟间嫉妒带来的损失。

In recent years,with the rise of the electricity industry and the increase of users′demand for electricity,the traditional power system no longer meets the users′demand for electricity consumption.While smart grid can handle the electricity demand of all parties well,facilitate the balance between supply and demand,and make the allocation of power resources more reasonable and efficient.As one essential research subject in smart grid,balancing power allotment based on demand response,has aroused a wide concern among scholars.In their studies,various new energy generation methods have been put forward to balance power allotment.Especially,the popularity of renewable energy generation system not only effectively reduces pollutant emissions,but also changes the traditional cooperation mode of power supply and demand,so as to provide many possibilities of improving the utility of power grid.Therefore,while studying the power generation of distributed power plants,this paper designs a penalty mechanism to reduce power loss or additional power generation costs by calculating the difference between the predicted electricity and the actual generated electricity of distributed power plants.In the designed penalty mechanism,the distributed power plants can exchange surplus electricity through cooperation to reduce penalties.At the same time,the alliance envy factor which is generated by the cooperation of distributed power plants is considered as well.Specially,the penalty mechanism proposed in this paper divides the power delivery into two stages,and calculates the difference between the predicted electricity and the actual generated electricity of distributed power plants for the implementation of penalty mechanism.The first stage is the preparation stage,when distributed power plants forecast the power generation within a certain period in the future according to their respective historical power generation data,and then the market operator collects the predicted power generation information and announces it to the power market for sale.So the power generation plan is determined according to the real-time market sale price and the electricity clearing price is predicted.The second stage is the day-ahead stage,when the market operator collects the actual power generation level of each distributed power plant,and takes the absolute value of the difference between the predicted power generation and the actual power generation as the measurement standard of punishment.If the predicted power generation is closer to the actual power generation,the unnecessary waste will be greatly reduced.Otherwise,more surplus power resources or insufficient supply may be generated,while the power plant bears the additional penalty for the waste of power resources.In response to this penalty,power plants can cooperate with each other to transfer excess power to power plants with insufficient power so as to reduce the punishment effect.On the other hand,the power plants will group members to form an alliance according to the standard of power shortage or surplus.The unit power transmission cost between each two is used as the index to sort the power transmission since power transmission will generate corresponding costs,then the power transmission is carried out in order to meet the power demand of each power plant.The above penalty mechanism not only minimize the cost of transmission,but also greatly reduce the revenue loss of power plant.In addition,based on the domestic electricity reform policies and the current situation of the electricity market,this paper introduces human behavior factors to the electricity market.That is,when multiple alliances exist at the same time,the party with lower alliance revenue will envy the party with higher alliance revenue,which is manifested in the negative growth of his revenue.In order to reduce the negative impact of the envy factor on revenue,the form of alliance between power plants will be changed to achieve respective ultimate utility maximization.Finally,the utility of each distributed power plant is allocated according to Shapley value method.In the simulation section,the penalty coefficient is analyzed and it seems that when distributed power plants cooperate,the change of penalty coefficient has a greater impact on the players in the alliance.If a member in the alliance is selected for electricity exchange due to electricity difference,he will have a positive feedback on the penalty coefficient,otherwise it will be a negative feedback.The analysis of the envy coefficient shows that,firstly,if the power plant chooses the alliance in which the total power difference of the alliance is opposite to the power difference of its own,the increase of envy coefficient is more beneficial to its utility growth.Secondly,when considering the envy between alliances,the envy coefficient will have a negative impact on the members of the unallied or low-utility alliances.To prevent this phenomenon,each power plant can join an alliance with higher total utility to avoid the occurrence of envy and utility lose.Thirdly,the envy factor has a significant impact on the members whose initial utility is at both ends of the alliance:If the member′s initial utility is the lowest,its utility will decrease as the envy coefficient increases,and vice versa.The members with average utility in the alliance is less affected by the envy factor,and the change of their utility is smaller than that of other members.The penalty mechanism proposed in this paper is of great significance to guide the practical application of cooperation decision-making of distributed power plants.