Loss-minimisation techniques applied to load-flow analysis form an effective means of obtaining a desirable voltage solution. In addition, the method offers improvement in convergence, which is especially important for networks which give rise to ill conditioned equations. The described method uses the nodal-impedance matrix and minimises that part of the total real-power losses which depends on the generator reactive-power input and on the transformer tap settings. The reactive-power input and voltages at the generator busbars, together with transformer tap settings, are calculated automatically by the program according to the criterion of minimum loss. In contrast to the methods previously employed, the off-nominal turns ratio of transformers is represented by a superimposed branch current and an injected nodal current. The transformers may be in-phase or phase-shifting. The paper shows how the minimisation procedures may be incorporated into the iterative process of a general load-flow program. A comparison with other methods is given by means of a sample-system study.