The invention relates to an optimal scheduling method of a nonlinear-programming cascade reservoir group based on a GAMS (general algebraic modeling system). The method comprises the following steps of: establishing an objective function; considering constraint conditions such as reservoir water balance constraints, related engineering constraints, related power generation constraints, other policy constraints, maximum constraint on minimum output, maximum constraints on guaranteed output, minimum constraints on differences between initial and final reservoir capacities in transition periods, and maximum constraints on assurance rates; and solving. According to the invention, runoff processes occurring in the past, artificially-generated runoff sequences and future runoff forecast processes are all taken as deterministic input processes of reservoir operation scheduling. The runoff description method can be used for intuitionally and visually reflecting the continuity and periodicity of runoff changes in a certain period, and indirectly reflecting the statistical properties of runoffs. According to given runoff data, the operation of solving is performed by virtue of a deterministic optimization model to obtain an ideal optimal operation process; and then, according to some criterion, a scheduling discipline function is identified from the optimal operation process and is used for guiding the operation of a hydropower system.