Modular parallel technology to increase capacity has advantages such as plug and play.This technology has become an important development direction of high power converters.The traditional centralized control needs to add auxiliary communication network,and the system cost increases significantly.Distributed control requires no independent communication network,but the multi-loop control structure is prone to low bandwidth and slow response. In order to solve the problems,we propose an improved distributed architecture segmented predictive current control (SPCC)method.Firstly,a unified zero-sequence model of multi-module parallel system is established,and the dynamic behavior of zero-sequence current is compared and analyzed.On this basis,thediscrete prediction model of the system is derived and the SPCC method is designed.In each control period,the voltage vector is decomposed into two parts, including the base voltage vector and the zero voltage vector,to meet the requirements of fixed switching frequency and improve the flexibility of the prediction architecture.Furthermore,the value function satisfying ac-straight axis current tracking and zero-sequence current suppression was constructed,and the principle ofoptimal voltage vector selection and duty cycle calculation were formulated.Finally,the proposed SPCC method is verified and analyzed based on a double 45 kW test prototype.The response time from no-load to full-load is 2 ms,the steady-state current THD is 3.1%and the power factor is greater than 0.99.At the same time,it has the advantages of fixed switching frequency,dispersion and no interconnection.