Chapter 10 Time Series Decomposition

Typical time series result from the interaction of regular deterministic and random causes. The deterministic regular causes can vary periodically (seasonally) and/or contain long-term trends. Random causes are often also called noise.

Decomposition breaks down time series into their components “trend”, “seasonality” and “randomness”. For further data analysis it is often necessary to decompose the trend and the seasonal component from the raw data. Therefore, a large part of the time series analyses deals with corresponding procedures. Sometimes the remaining Randomness is of interest and sometimes a detrended series as in chapter 13.4.

Trend Component
The trend component of a time series refers to the general direction in which the time series moves in the long term. Time series can have a positive or negative trend, or no trend, but they can also have no trend. A trend is present when the data show a continuously rising and/or falling direction.

Seasonal Component
The seasonal component for time series data refers to their tendency to rise and fall with constant frequency. Seasonality occurs over a fixed and known period of time (e.g. the quarter of the year, the month or the day of the week).

Random/Remainder/Irregular Component
The rest is what remains of the time series data after its trend and seasonal components have been removed. It is the random fluctuation in the time series data that cannot be explained by the above components.

For energy data analysis, each of the three components mentioned above may be of interest, depending on the case.

This chapter shows how a time series can be decomposed and presented interactively.

An exemplary decomposition of an outside temperature time series over 7 days:

Figure 10.1: Time Series Decomposition Outside Temperature over 7 days