There are a few issues at play here.
First is the concept of a synchronized electrical grid, using the "infinite grid" approximation, in most cases a reduction in load on an electrical system will be imperceptibly distributed across all the generators in the grid. The load that a specific power plant sees is a managed load (managed in the context of the entire grid) and the system may respond to a large decrease in a number of ways outisde of the realm of a single generating station. If half of the entire load on the national grid drops out suddenly then you have bigger issues to deal with than a single coal generating plant going offline.
A generator will have a speed governor and a droop rating for the given prime mover (ratio of speed at full load vs no load) . A decrease in generator load will in general increase the rotational speed of generator from steady state, the speed governor will command the power plant to produce less torque (less heat or less steam into turbine) and decrease the rotational speed. In a synchronized grid setting this process is "clamped" - meaning that if the generator represents a small fraction of the grid power it will be unable to increase its frequency to the new steady state and will go out of sync with the grid if this doesn't happen reasonably quickly (or if it is a large fraction of power generated, take the whole grid out of sync)
What this means is if the load on your generator is suddenly decreased (say local sector load dump) it must compensate by decreasing prime mover torque output or dump power into a quick turn-on load to avoid risk losing synchronization with the grid
That being said power =/= energy. If your governor and prime mover have quick response (say seconds) in reducing generator output then 1s at 50MW is 50MJ, or approximately 13 W-hrs of energy. This is not a huge amount of energy in a power plant context and can easily be dissipated as heat through dump loads. If you need to divert 50MW for hours, this is a large amount of energy that will be better served using energy storage. You will need to determine the dynamic properties of the coal plant and prime mover to determine the best course of action.
E.G. if the turbine has a steam reservoir then the heat output of the coal is buffered and you are more likely to have quick control over prime mover output independent of the heat produced by the coal fire. Given enough physical energy storage (as heat in the reservoir) to ride out the load dump or buffer the turn down delay then excess power loads of some sort will be enough depending on the expected response. If the requirement is that "coal burner operates at louvres=100% for 100% of the time" Then yes, you need more energy storage, the most efficient being upstream of the generator (heat diversion) because conversion to mechanical force, then EMF, then any electrical storage then back to mechanical and EMF will be much less efficient overall, but I'm sure that this has been solved worldwide with much less handwaving and more hard economic numbers so you will have to read the literature.