Power flow studies are very important and are essential for fault analysis and stability analysis. Conventional Newton–Raphson-based power flow analysis is for fixed loads and generations. However, load and generation powers are uncertain due to factors like forecasting error, sudden outages, etc. Monte Carlo was widely used for power flow analysis with uncertainty. Due to excessive computational time involved in Monte Carlo, affine arithmetic-based power flow analysis was introduced. In the present work, a modified affine arithmetic-based power flow analysis is proposed. The proposed method needs less computational memory and gives tighter bounds on voltage magnitudes and angles of the system buses for a given uncertainty range of load and generation powers as compared to existing affine arithmetic-based power flow analysis. The proposed method is implemented on IEEE 30 and IEEE 118 bus systems and detailed simulation results are presented. The results are compared with Monte Carlo simulations as well as existing affine arithmetic-based power flow analysis. The results show that the proposed modified affine arithmetic-based power flow analysis yields more accurate solution intervals for voltage magnitude, voltage angles, real, and reactive power flows as compared to existing affine arithmetic-based power flow analysis. © 2018, Copyright © Taylor & Francis Group, LLC.