In power systems with a significant share of hydropower, planning the operation of hydroelectric power plants is of great importance. The lack of knowledge and the stochastic nature of future inflows are the main problem. The goal of planning is to find an optimal way of using the available amount of water while trying to satisfy numerous constraints. In an optimization calculation, natural inflows at a certain location can be represented deterministically, in which inflows are treated as known quantities, or stochastically, in which stochastic characteristics are already incorporated into the calculation in the input data model. The goal of the paper is to evaluate a suitable stochastic model for generating inflows at a certain location. Two stochastic models were tested that treat inflows in different time periods as mutually dependent random variables described by a first-order autoregressive Markov model. One model is used if the previous inflows at that location are subject to a normal distribution, and the other if the previous inflows at that location are subject to a log-normal distribution. After developing the program and simulating a series of inflows using both models, various statistical characteristics were examined with the generated inflows: mean value, standard deviation, inflow correlation factors in consecutive time periods, etc. The generated inflows retained all the examined characteristics as the actual inflows measured at that location over a series of years. Both examined data models were assessed as suitable for the needs of medium- and long-term simulations and optimization of hydroelectric power plant operations.