I am doing some data analysis where I look at electricity demand over the course of a day, but need to separate the intra-day (constant and periodic) components from daily changes (assumed linear). At its very simplest it boils down to a scenario of a function of the form $y = a t + b \sin n t + c \cos n t$ over the interval $t \in [0,2\pi)$.

The question I have is are there any well recognised algorithms for removing the $t$ component? The problem I am having is that $t$ is not orthogonal to $\sin t$ over the interval $t \in [0, 2\pi)$, so any simple algorithm like doing a matrix regression using QR results in perdioic basis waveforms that are no longer sinusoids.

Are there any special algorithms for removing a ramp from a sinusoid without distorting them? In general I am not doing a full fourier transform, as a day's data has 48 samples but I am in general only interested in the components with at least a 2 hour long cycle.


If $n$ is large enough, linear regression should do the trick. In other words, you can jut consider the residuals of the linear regression.

  • $\begingroup$ True, but I want it to be accurate for any $n$, in particular $n = 1$. $\endgroup$ – bjem Jul 16 '15 at 5:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.