In the background of the new power reform, correct and reasonable assessment of the credit risk of power usersand early warning are important factors that restrict the further development of power supply enterprises. In order to carry out creditevaluation of power users and help power supply enterprises to balance the risk level of power trading, based on random forest algorithm, the evaluation principles and indicators of power users arestudied accroding to the evaluation indicators of credit risk in power market. The early warning model of power users evaluates thecredit risk of power users from enterprise market risk, financialrisk, credit risk, operational risk and settlement management risk.It is expected to improve the trading mechanism of electricity market and promote the rational and efficient allocation of resources,help power industry develope healthily under the protection of credit system.