You can use the calculator [on this website](http://bikecalculator.com/wattsMetric.html), which will give you a reasonable estimate if you know the average grade of the hill, the day's temperature, and the wind speed/direction (probably not so relevant on a hill). A similar calculator is [here](http://sportech.online.fr/sptc_idx.php?pge=spen_esy.html) so you can compare two methods.
 
The website http://www.cyclingpowermodels.com has a host of information about power models, including the following excerpt. I couldn't find a power calculator on there though (only the opposite).
 
> **Model Validation**
> 
> Two key questions in the application of any model to the analysis of
> cycling must be "is it accurate?" and "what are the assumptions?"
> 
> Models of the relationship between cycling power and speed have been
> around for a long time and rely on the physical principles in Newtons
> Laws of Motion. The principal model of cycling power and speed used
> throughout this site is an implementation of that proposed in
> Validation of a Mathematical Model for Road Cycling Power  which
> appeared in the Journal of Applied Biomechanics in 1998. This
> publication demonstrated the completeness and validity of the model by
> comparison of model predicted and observed power values. The model
> calculates the power a cyclist would have to produce in order to
> achieve a certain speed on a certain course, taking account of key
> physical and environmental parameters. In places this model is used to
> compute speed, time or the value of another parameter given a
> specified power.
> 
> The performance of any model is only as good as the accuracy of it's
> inputs which is why we often go into great detail measuring or
> estimating major variables such as air density, wind and aerodynamic
> drag. Any asumptions or modelling approaches will generally be
> outlined. To some extent the use of consistent power models in the
> field derivation of aerodynamic drag measurements (i.e. field testing
> of CdA) can improve the reliability of the models when used with that
> input.
> 
> In practise we have found theoretical values of a cyclists ride time,
> given a specified power and good parameters inputs, to fall
> consistently within +/-5% of actual ride time and frequently within
> +/-2%. In the context of the stated accuracy of most cycling power meters at +/-2% we have great belief in the application of physical
> models to the analysis of cycling events and, more importantly, the
> analytical power provded to the rider or coach. The more we use these
> models the more confidence we have in them - if you have used them
> feel free to let us know your findings.

Note that gear ratio and cadence is not necessary for the calculation (you can produce the same power output with lower gears and faster cadence or vice versa).