It sounds like you're asking about a switching power supply? At higher frequencies, all else equal, the ripple will be reduced, because you won't be depleting the charge in the capacitors as much before they are recharged, which means you can reduce the capacitance for same amount of ripple as you would when operating at a slower switching frequency.

I think the main issue becomes power wasted due to stray capacitance, in traces and passive components, as the switching frequency goes higher and higher. For example, inductors have stray capacitance. At low frequency, that may be insignificant. At very high switching frequency, the inductor can become bypassed, as current essentially finds a short circuit through the stray capacitance.

My point here is that, for different switching frequencies we redesign the inductors and capacitors. At higher switching frequencies, the inductor and capacitor size gets reduced. Other things will remain the same. The power converter behavior will remain the same. i.e. It will give the same output at every frequency. So for this process, the inductors and capacitors have to store the same amount of energy at every frequency. The reduction in their size will have impact on their current and voltage as far as energy is concerned. i.e. 0.5LI^2 and 0.5 CV^2.

The energy stored by these devices will change with switching frequency and it will be too small for higher switching frequency. But you are considering it for fundament frequency component of voltage and current. However, voltage and current should be taken for ripple frequency i.e. delta_I and delta_V. 

You do not have to store the same amount of energy when frequency is increased. Every time you double the frequency you just have to store half of the energy you stored before for same performance (it is not this straight forward because of leakage and etc. but more or less like that). 

In my view point, by increasing the frequency the component size gets reduced and the operation voltage reduced. Since the energy stored depends on voltage, it stores reduced energy storage but the performance will not be reduced.

If the power output is the same - irrespective of the frequency - the 'stress' for the power components (L, C) is decreasing (proper dimensioning assumed) with increasing frequency as the "bits" of power transferred during the active cycle decrease. This is expressed in lower ripple. The stress for these components is proportional to the square of the ripple while the some DC component is only relevant for the inductor (especially when using CCM.

What's increasing with the frequency are the switching losses - adding stress to the active componejts.

In higher switching frequency the parasitics of the LC components and the switching FET will influence more then in lower frequencies. you will have to look carefully on the frequency  behavior of these components.

Dan

For passive components, the main thing to watch is the capacitors. If you go through the math. you will find that they will not only have a storage charge and discharge power of 1/2CV^2F, but also, an ESR dissipation of 1/2CV^2F, . F is the switching frequency and ESR is the effective series resistance of the capacitor. As you can see, the dissipation increases directly with F. This is of particular concern with electrolytic capacitors when their electrolyte heats up too much and the capacitor fails. To answer your original question, if the energy storage remains the same then the capacitor will have more power loss and dissipate more heat at higher switching frequencies. This should be included in a stress analysis for the design.

 I agree with Thirumurugan Ra that the reduced charging and discharging time at high frequencies reduces the energy stress on the smaller capacitor and inductor. This can be easily visualized in a parallel plate capacitor with a smaller area for the plates and/ or a longer length of separation between them as a more relaxed capacitor with less charge concentration on the plates and/or a lower field between the plates for the same voltage. The same logic can be extended to an inductor or solenoid in terms of the area of cross section and length of the iron (better for visualization) as being driven by a lower magnetic field and/ or having a lesser flux density. This is a simplistic presentation of the idea, no doubt but the visualization helps younsters. 

In order to quantify the problem, let us assume that the passive elements are those reactive elements such as Inductors and capacitors used to perform voltage conversion and filtering operation. As the frequency is up scaled by factor m then the inductor will be scaled by 1/m , and also the capacitor will be scaled by 1/m to scale the cut frequency of the filter by m. This can be get out from the fact that the resonance frequency w0^2= 1/LC. So, it is clear from this relation that if wo is scaled by m , both L and C are scaled by 1/m as i assumed in the beginning of the post.

Now let us assume that the supply Vs operates at the same Voltage, to supply the same power to a resistive  load RL. So the problem is now defined. One can calculates the current in the inductance assuming that the whole voltage will be dropped on it , then 

The current in the inductor IL= Vs/ jwLi where Li initial inductance. Scaling the frequency by m then IL scaled= Vs/ jwm Li /m= IL, so the current will not change by the scaling. 

The power dissipation will become= IL^2 scaled Rl with Rl is self-resistance of the coil.

The self resistance of the coil is proportional to its number of turns N. It is known that the inductance L is proportional to N^2, so  Rl will be scaled by 1/ sq. root of m.,

So the power dissipation in the inductor will be scaled by 1/ sq. root of m under the given assumptions. If the resistance is made initially small there will be no appreciable temperature rise by scaling the inductor. The power density per volume and the surface area of the component can be designed to keep the temperature the same.

As for the capacitor, as it is nearly ideal its losses will be small and the temperature rise by frequency scaling may be negligible.

More rigorous analysis is required to quantify the problems but according to these estimations it is not very serious.

Best wishes