The conventional design of HVAC-DC system controllers is based on model linearisation at operating points, hence transient performance cannot be guaranteed. A minimum variance adaptive control scheme is developed for improving the generator performance and hence power system stability. The controller supplies the excitation system and both terminals of a DC link with control signals derived from a control strategy which minimises the variance between system outputs and outputs of a linear predictor. Results obtained from a detailed nonlinear simulation of a two-machine AC-DC system are presented.