The optimal operation of reservoir group is a multi-constrained and high-dimensional optimization problem. For this problem, this paper couples the standard pelican optimization algorithm with the adaptive ε-constraint method, and further improves the algorithm's optimization-seeking perfor-mance through the good point set initialization population, reverse differential evolution and op-timal individual adaptive strategy, and proposes an improved pelican algorithm (ε-IPOA) based on the adaptive ε method. The performance of the algorithm is tested by eight constraint test functions to find the optimal ability and solve the constrained optimization problem, and the results show that the algorithm has a strong ability to find the optimal and stable performance. In this paper, we select Sanmenxia and Xiaolangdi reservoirs as the research objects, establish the maximum peak-cutting model of terrace reservoirs, apply the ε-IPOA algorithm to solve the model, and compare it with the ε-POA and ε-DE algorithms, the results show that the ε-IPOA algorithm is better than other algo-rithms, the peak clipping rate of Huayuankou control point solved by ε-IPOA algorithm reached 44%, and other algorithms do not find effective solutions meeting the constraint conditions. This paper provides a new idea for solving the problem of flood control optimal operation of cascade reservoirs.