To provide a fuel injection learning control device for an internal combustion engine capable of reflecting this time final learning value calculated in learning control in a calculation of a target driving current value impressed on a solenoid coil of SCV without performing excessive and complicated control.

In the learning control for learning and correcting the functional deterioration (performance reduction) of the SCV by an individual difference and a change with the lapse of time of the SCV, this time temporary learning value is calculated to be reflected in the calculation of the target driving current value impressed on the solenoid coil of the SCV, and afterwards, a value of adding last time learning value and this time learning value is calculated as this time final learning value. However, in this time operation switched to an ordinary control condition from a learning control condition, this time final learning value calculated in the learning control is not immediately reflected in the calculation of the target driving current value impressed on the solenoid coil of the SCV, and when stopping an engine, this time final learning value is only stored in a memory such as EEPROM instead of the last time learning value.