The power generated by a hydroelectric plant depends on the penstock head loss, the turbine efficiency, and the generator efficiency among other factors. Initially, the functions related to these variables are provided by the equipment manufacturer; however, over time they change as the plant ages. This paper presents a methodology to adjust an efficiency function for each generating unit based on measured data. It is applied using two optimization methods: the generalized reduced gradient and the evolutionary algorithm. A case study with data from a large Brazilian hydroelectric plant demonstrates how the methodology can be employed and compares the performance of the optimization tools. A comparison metric is used to show that the optimal unit efficiency function significantly improves the performance of simulation models to reproduce observed data and better describe the actual operation of the hydroelectric plant. DOI: 10.1061/(ASCE)EY.1943-7897.0000074. (C) 2012 American Society of Civil Engineers.