In recent years there have been some major developments in the space industry with the introduction of fully electrically propelled satellites. All-electric propulsion enables smaller and lighter satellite design that reduces the launch costs by a big margin when compared to traditional satellites with chemical propulsion. Due to its efficient propellant usage, solar electric propulsion (SEP) has now become an attractive option for low thrust orbit raising to geosynchronous equatorial orbit (GEO). However, the key issue with SEP is its long transfer time spanning several months. Moreover, this long transit exposes the solar arrays of the satellite to the highly energized trapped particles of the Van Allen radiation belts. The charged particles have serious effects on the mission lifetime because the solar cells generate power for all operations and may get damaged by the radiation exposure. Thus, it has become very important to characterize the power loss due to radiation damage. This thesis proposes the use of an artificial neural network (ANN) based framework for predicting the power loss of all-electric satellites during low thrust orbit-raising through the Van Allen belts. The developed networks can be beneficial in computing thrust availability of the satellite within low-thrust mission design tools that compute optimal orbit-raising trajectory for the satellite. In this study, a radiation dataset for the trapped particles is created using the combination of AE9/AP9 radiation tool and SPENVIS software developed by the European Space Agency (ESA). The dataset is then used for training of an ANN, which predicts the loss of power during orbit raising maneuvers. The application of the network to a sequential low-thrust orbit-raising solver is demonstrated through numerical simulations for orbit-raising starting from different geosynchronous transfer orbits. The specific solver was chosen for its ability to compute low-thrust orbit raising trajectories in a fast and robust manner. It is observed that the satellites output power can deteriorate approximately by 15%-25% to reach its final destination of GEO. When compared to conventional methods the proposed technique provides comparable results for solar array degradation; moreover, the rapid computations using ANN preserves the fast and robust nature of the solver.