The operating characteristics of a fluid machine are represented by a characteristic curve that expresses the correlation between the flow rate and pressure of the working fluid. Since the power at the system's output end can be inferred using the flow rate and pressure information, the input power information is collected together, and the operation efficiency information is plotted on the characteristic curve for use. Based on the information, industrial sites are trying to promote the energy-efficient operation, and related studies in academia are also considering optimal control methods for fluid machineries based on operating efficiency. However, since efficiency simply considers the power ratio, there is a blind point that it does not reflect the operating characteristics of fluid machines according to load conditions. That is, the operating condition at the optimal efficiency point cannot be a necessary and sufficient condition for minimum power consumption. In addition, although data must be obtained in various operating areas to show 2D efficiency information on a characteristic curve, there is a practical problem that it may be insufficient due to environmental constraints. Accordingly, this paper proposes a method for constructing a surrogate model for input power using limited data for a 22 kW pump system. In addition, the effectiveness of the input power-based control method compared to the existing operation efficiency-based operation control method was verified through simulation.