Couple of factors at play here, but the short answer is ... it might not have been your imagination. Or it may have. It's hard to say without actually measuring the speed (cell phone with a gps app will do this) and comparing speeds of the same train at different times of day. Comparing rides from one train to the next on the same ride can yield slight differences in speed.
There are actually many examples of people asking this question online (even a few on Dis).
Considering only rides with rolling trains on tracks (aka roller coasters) there are a few mechanical aspects that can affect the ride based on temperature.
- Every wheel is supported by a race of bearings. These bearings are usually sealed and greased. When the grease is cold it has a higher viscosity and could slow the train somewhat. This effect would be minor and by the time evening comes around, the ride in use would create enough frictive heat to nullify this effect. But a train put fresh into service and the first runs of the day (which probably run empty) might be affected. I would say this effect, if measurable at all, isn't a significant factor.
- Pneumatic braking and speed controls. We all know the braking that kicks in at the end. But often there are speed control braks at key points throughout the ride. The pressure tanks for these mechanisms are often in full sun and a difference of temperature could move the braking pressure from one end of their acceptable range to the other. This would cause a few milliseconds difference in the amount of braking applied which could translate to a speed difference of a MPH or two by some point along the route of that coaster. Similar to point #2, I doubt this is causing a noticeable effect.
- Steel track rides with significant exposure to the sun do exhibit a measurable degree of expansion. Well within the tolerances of the ride, the track at its hottest would be gripped tighter by the rolling trucks of wheels that support each train car. Simply put, the expansion of the track increases the friction on those wheel bearings. It may not seem like much, but this increase in friction creates a cumulative effect along the entire route of the ride. It's conceivable to me that at some point near the end of the ride there might be enough difference in speed that one would notice it.
But we're not done yet...
Another effect that is often brought up to explain this phenomenon is that riding a roller coaster in the dark feels faster because the rider has a much diminished depth perception. I've run video simulations of first person point of view motion (the view of a road from inside a car) and altering things like field of view and depth of field produce radically different feelings of how fast the car (in this case) is traveling.
So ... who's gonna gather some data? Apps designed for tracking your run or bike ride would work. I'm setting up one called Strava for my DLR visit tomorrow.