This study refers to developing an electric-power distribution system  with optimal/suboptimal load-sharing in the complex and expanding metro power-grid infrastructure.  That is, the relevant exercise is to indicate a smart forecasting strategy on optimal/suboptimal power-distribution to consumers served by a smart-grid utility.  An artificial neural network (ANN) is employed to model the said optimal power-distribution between generating sources and distribution centers.  A compatible architecture of the test ANN with ad hoc suites of training/prediction schedules is indicated thereof. Pertinent exercise is to determine smartly the power supported on each transmission-line  between generating to distribution-nodes.  Further, a “smart” decision protocol prescribing  the constraint that no transmission-line carries in excess of a desired load.  An algorithm is developed to implement the prescribed constraint via the test ANN; and, each value of the load  shared by each distribution-line  (meeting the power-demand of the consumers) is elucidated from the ANN output. The test ANN includes the use of a traditional multilayer architecture with feed-forward and backpropagation techniques; and,  a fast convergence algorithm (deduced in terms of eigenvalues of a Hessian matrix associated with the input data) is adopted. Further, a novel method based on information-theoretic heuristics (in Shannon’s sense) is invoked towards model specifications. Lastly, the study results are discussed with exemplified computations using appropriate field data.