For Generator Protection, what protection scheme would lessen my Incident Energy? This is my understanding of Generators; typically emergency generators produce a much lower short circuit current than the normal source. However, this allows the buildup of incident energy over time. How do I go about mitigating this incident energy build up since there is no upstream protective device? How do I reduce the fault-clearing time thus lessening the arc flash hazards?

Classic problem. First solution is this: don't panic. There are multiple options simply because every case is unique. Arc flash mitigation is not always easy or inexpensive to accomplish but there are always solutions.

Essentially if you look at your protective devices, you have 3 regions of operation:
1. The area below the curve in the "normal" operating region where it won't trip. If you have an arcing fault here, it remains undetected and we typically model this with a 2 second "timeout" on the incident energy. The lower the current, the lower the incident energy.
2. The area within the "instantaneous" operating region of the device. At this point the device has a finite opening time. The lower the current, the lower the incident energy. This is a region where current limiting can be a benefit whether current limiting fuses or circuit breakers, but note that it becomes a detriment for case #3 below because it increases the tripping time further.
3. The "inverse time" region of the device. This is where some very unexpected things happen. As the fault current decreases the arc power decreases as expected, but also the clearing time of the protective device is increasing and the rate that it is increasing is an inverse whereas the decrease in arc power is mostly linear, so the net result is that as fault current decreases, incident energy actually increases.

The first thing to understand is...what condition are you in (1, 2, 3) above, or more complicated, are you switching from one region to another? Because in cases 1 and 2, the decreased short circuit current of the generator is actually beneficial in lowering incident energy but if you are in the inverse time region, it actually increases incident energy.

There are a couple ways to combat this, some simple and some not so simple. The first one is that the generator has its own protection to begin with and this should provide adequate protection or can usually be inexpensively modified to provide adequate protection (fast enough tripping) on the bus that the generator is connected to. Downstream overcurrent devices though may still not provide adequate protection at that point. The choices will be either to reduce the size of the downstream devices to make it work, or provide a dual-setting arrangement, or provide some other means of tripping. For instance if everything is close enough together that you can do this, zone selective interlocking allows essentially all breakers that are ZSI-interconnected to trip on instantaneous overcurrent so the inverse time region is no longer required for selectivity at all. If you have electronic breakers already this may be an option but if you don't the upgrade can be pretty expensive.

Finally there is the less-than-desirable but still effective option of simply posting two different labels, one showing the incident energy on backup generator power and one without. Even if the incident energy becomes a very large number, it may mean that you simply don't work on the equipment when the emergency backup generator is running and either shut it off or only do energized work on primary power, or you use a different circuit breaker to isolate equipment compared to normal operation such as the breaker connected to the generator itself. If you only run the backup generator once in a while, this is not necessarily a terrible option.

Just some ideas...really each case is unique but there are a half dozen or so strategies that are almost universally applicable. It's just a matter of checking each one and picking the best option.

To add to Paul's comment, the protection coordination exercise can become complicated for near faults to the generator, where the fault currents are varying in accordance with the generator decrement characteristic and the changing generator excitation in response. In plotting the fault current with time on the protection characteristic, I have had cases where available fault current drops below the IDMT characteristic, allowing the protection to reset before the fault is cleared.