Understanding and predicting the performance of solvent drives and remediation of contaminated aquifers in heterogeneous reservoirs is of great importance to the petroleum and environmental industries. In this paper, a general method to scale flow through heterogeneous reservoirs is presented for a miscible displacement of oil by a solvent. Results show that scaling miscible displacements in a two-dimensional, heterogeneous, anisotropic vertical cross section requires the matching of 13 dimensionless scaling groups. These groups were derived using a general procedure of inspectional analysis. A detailed numerical sensitivity study was performed to reveal the relationship between the scaling groups and the fractional oil recovery of miscible displacements in heterogeneous reservoirs. This relationship was then mapped using an artificial neural network, which can be used as a quick prediction tool for the fractional oil recovery for any combinations of the scaling groups, thus eliminating the need for the expensive fine-mesh simulations. These results have potential applications in modeling miscible displacements and in the scaling of laboratory displacements to field conditions.