After replacing aging or faulty equipment, thermal power plants encounter the need for simultaneous operation of new and old equipment. Following the replacement of the one step-up transformer in multi-machine power plant non-uniform distribution of reactive powers and generators’ terminal voltages occurred due to different step-up transformer transmission ratio. Uneven distribution results in higher generator heating for the generator operating at a higher apparent power, non-uniform distribution of stators’ voltages in p.u. that results in varying distances to the allowable voltage limits, reduced reactive reserves, etc. In the process of allocating reactive powers among generators, the goal is to achieve a balanced distribution of apparent powers. This ensures uniform heating of the generators and avoids operating them close to their thermal and voltage limits, where zones of accelerated generator aging are identified. Minimizing the voltage difference at the generators’ terminals ensures an even voltage stress and an equidistant operating point of the generator from voltage limits. The third criteria, maintaining the maximum reactive reserve during voltage increase and decrease at the connection point ensures best power system reactive support. These requirements include ensuring that the difference in the generators’ terminal voltages is not too significant, that the generated reactive powers on the generators are similar, and that the generated reactive powers approach their permissible limits gradually. Due to the arbitrary nature of these requirements (not too significant, similar, gradually), it is difficult to formulate the problem mathematically. However, constrained optimization is focused on determining a feasible solution rather than an optimal one. Therefore, the problem of allocating reactive powers, as described above, is best addressed through constrained optimization. This method requires defining constraints, variables, and an objective function and then finding a feasible distribution that satisfies both the soft requirements and hard limits.This paper presents a series of calculations performed to analyze the optimal allocation of reactive powers among the generators using an optimization procedure with constraints. Several optimization methods can be used for computing optimization function minimum or maximum. Understanding these methods is crucial for selecting the most appropriate approach for the defined problem and correctly interpreting the usefulness of the obtained results. In paper the evolutionary method is chosen. Several optimization functions were tested. Desired optimal allocation is computed. The impact of the voltage at the connection point was taken into account for the selected optimization criterion. Calculations were performed, and the optimal tap position was selected. However, it has been shown that the evolutionary method's duration depends on the computational time limit, making it unsuitable for automatic allocation of reactive powers in real-time. Automatic allocation of reactive powers ensures even loading of generators. It also takes into account the available reactive reserve at the power plant buses to cover the dynamics of renewable sources. As the development of allocation algorithms becomes more complex, it will be necessary to combine off-line algorithms from the family of artificial intelligence algorithms to create fast adaptive on-line control algorithms.