Abstract: With the deepening of the reform of electric power system, the consumers gradually occupy an important position in the electricity market. Research on electricity consumption in the market environment is beneficial to the optimization of electricity market mechanism and lay the foundation for the data operation of the market. It also has positive effect on the security and stability of power system. According to the response mechanism of different consumers to the electricity price and demand response policies in the electricity market environment, the paper put forward a market behavior evaluation index system from the aspects of electricity purchasing potential, electricity price sensitivity, demand response potential and demand response sensitivity, and provided quantitative models for each index. Based on the indexes for the existing consumer samples, the initial consumers were clustered through the affinity propagation algorithm. The clustering result was defined as the initial classes and a distance threshold was initialized. The indexes of each sample and the corresponding class were used as the input and output to train LVQ (Learning Vector Quantization) neural networks. The new sample indexes were input into the trained network to identify. When the distance between the sample and the identification result exceeded the threshold, the sample was considered as a special one which subordinate to a new class. New classes were trained from the special samples and the LVQ neural network was updated with the training result to achieve the effect of self-adaptation identification. The example shows that the identification result of the self-adaptive identification model is accurate and the model can effectively identify new market behavior and update the library of consumer types.