Aiming at the problems of single power optimization in actual industrial parks without considering environmental factors, a method for optimal power allocation to industrial park users based on carbon emissions is proposed. Firstly, a multi-objective power optimization model is established, which includes the economic benefit objective of comprehensive power generation cost and the environmental benefit objectives of peak-to-valley difference and carbon emission. Secondly, the improved particle swarm algorithm is used for multi-objective optimization solving, combining the traditional particle swarm algorithm with the immunity principle, replacing subjective weights with objective assignments, and avoiding the local optimum problem. Finally, the load data of some users in the industrial park are brought in and the data indicators before and after optimization of the park are compared and analysed.