To control wind turbine fatigue and optimize the fatigue distribution for offshore wind farm, a control network model is proposed based on Multi-Agent theory. A typical model of large-scale offshore wind farm is described. Power fatigue of individual wind turbine is defined. In offshore wind farm, a fatigue distribution estimation approach is studied. The Multi-Agent network is modeled based on the current communication network. All wind turbines act as independent agents, and can self-organize into wind power delivery groups adaptively according to the real wind direction. A solution for automatic agent controller synthesis using Uppaal Tiga and Simulink is presented. The fatigue distribution can be optimized to prolong the service time of a wind farm. This approach can also save some of the maintenance cost due to the expensive logistic visit by helicopters. Finally, a typical simulation result illustrates the feasibility of this multi-agent model.