Within a hybrid electric vehicle, given a power request initiated by pedal actuation, a portion of overall power may be generated by fuel within an internal combustion engine, and a portion of power may be taken from or stored within a battery via an e-machine. Generally speaking, power taken from a vehicle battery must eventually be recharged at a later time. Recharge energy typically comes ultimately from engine generated power (and hence from fuel), or from recovered braking energy. A hybrid electric vehicle control system attempts to identify when to use each type of power, i.e., battery or engine power, in order to minimize overall fuel consumption. In order to most efficiently utilize battery and fuel generated power, many HEV control strategies utilize a concept wherein battery power is converted to a scaled fueling rate. When battery power is used to propel the vehicle and hence is positive, the scale factor for battery power is chosen to produce a fueling rate estimate that, as nearly as possible, predicts the fuel rate required to recharge battery power during future vehicle operations. On the other hand, when the battery is recharging and hence battery power is negative, the scale factor applied to battery power is chosen to reflect fuel savings that will occur in the future, when battery power displaces engine power for vehicle propulsion. The latter case of scaled battery power produces a negative fueling rate to reflect the fact that a portion of the engine fueling rate makes power for later use. When the estimated fueling rate associated with battery power use is added to the always-positive fueling rate of the internal combustion engine, a so-called equivalent fueling rate is generated. A power split for maximizing fueling efficiency is then chosen by minimizing the instantaneous equivalent fueling rate. This paper will present a novel method for generating scale factors that convert battery power to equivalent units of fuel. Optimized power splits for specific vehicle cycles calculated using dynamic programming tools provide data for calculating scale factors. A neural net is trained to predict appropriate scale factors based on training data generated by dynamic programming. Input to the trained neural net is recent vehicle speed and battery state-of-charge history, and the output is scale factors for calculating an equivalent fueling rate.