A disturbance reduction scheme for linear systems with time delays and modeling uncertainties is presented in this paper. Unlike other disturbance rejection methods, the proposed scheme does not require information about unknown disturbance frequencies. The linear systems in this study are modeled to be nominally stable, minimum phase and relative degree one systems. The control structure is based on Astrom's modified Smith predictor with the proposed scheme consisted of an input disturbance reduction controller (IDRC) and a residual disturbance reduction controller (RDRC). The IDRC using an artificial neural network (ANN) is proposed to reduce an unknown input disturbance including unknown load disturbances and modeling uncertainties in both stable and unstable systems. The ANN can approximate appropriately a product of an inverse of a time delay and a nonnegative gain in the IDRC. In addition, the undesired responses caused by residual disturbances and residual modeling uncertainties are suppressed by the RDRC. Simulation results show the effectiveness of the presented disturbance reduction scheme for linear delay systems with modeling uncertainties, subjected to periodic unknown load disturbances.