The invention discloses an optimized distribution method for power of a thermoelectric combined energy storage power station, comprising the following steps: collecting relevant parameters of the thermoelectric combined energy storage power station; using the relevant parameters as index vectors of the power station; using the preset division limit of the state of charge, the power storage power coefficient, the heat storage power coefficient and the like of the energy storage battery as judgment conditions of a random forest neural network; defining the working state class of the power station; calculating whether the index vectors exceed a safe operation range under each working state class, calculating the probability of exceeding the safe operation range in each working state, and reserving the combination with the minimum probability as a training sample; using the training sample to generate and optimize the random forest neural network, taking the state quantity of the energy storage system in a future period of time as an input value, and outputting the working mode of the thermoelectric combined energy storage power station through the random forest neural network, therebysolving the demand problem of a control strategy for distributing power of an energy storage battery and a heat storage system in a future ultra-short-term power generation trend.