Abstract: Under the environment of power market, the factors of electricity price are more complex, while the flexibility of electricity price is greatly improved. From the perspective of market operation agencies and regulatory agencies, how to identify electricity price anomaly and explore the reasons under the complex market environment and the condition of incomplete data information, is one of the key issues to promote the orderly operation of power market and ensure the reasonable interests of power customers. A feature extraction and anomaly identification method of power customer price is established in this paper. First of all, electricity price feature vector is constructed, and the dimension of feature vector is reduced based on spectral clustering algorithm. Then, typical electricity price characteristics are extracted as the basic standard for determining whether electricity price is abnormal. Next, the similarity between each power customer and typical electricity price characteristics is calculated one by one. Finally, abnormal electricity price is identified in two stages. Abnormal reasons are positioned from two aspects of electricity consumption behavior and trading behavior preliminary and rapidly , and then further identified in depth on this basis. As the case analysis shown, this method can extract typical electricity price features and identify anomaly quickly and effectively. The reasons for electricity price anomaly are further analyzed from two aspects of electricity consumption behaviors and trading behaviors. Improvement measures are put forward accordingly.

Key words: power market, electricity price, spectral clustering, feature extraction, anomaly identification