面向混合增强智能的省级现货市场报价样本集增强及市场力识别技术OA北大核心

With the advancement of electric power spot market reform,there is an urgent need to conduct market power identification on the quotation behavior of provincial spot market participants to encourage healthy competition.Considering that in the initial stage of the market,the number of samples derived from the spot market practice in which market members exercise market power to obtain excess profits is relatively small compared to the samples from the normal trading behavior,there is a sample imbalance problem.In this paper,first of all,the market power label tagging was labeled based on experts'experience.In this paper,the samples of spot market practice with market power labels based on experts'experience were labeled and used to train the Random Forest algorithm for market power identification,and then enhance the sample imbalance to improve the accuracy of market power identification by using the sample sampling method.In addition,it is considered that as the market matures,the criteria for judging market power are not static,and in order to cope with the changing criteria for market power,the"human intelligence"and the"machine intelligence"were combined in this paper.The results of the algorithm show that after the market power label is changed by"human intelligence",and"machine intelligence"can still recognize the market power samples effectively,which reflects the advantages of the hybrid augmented intelligence proposed in this paper.