I have a time series data of hourly consumption of electricity for 720 data points and I want to make a forecast for 25 time points after the 720th time point. I am using the software R for this purpose.

First I have decompose the series into three parts:
(i) Trend (by method of Moving Average or MA)
(ii) Seasonal
and (iii) Residual or Error part,
The error component was stationary, so I used ARIMA model on that error component and predicted its value for the next 25 time points.

Now the challenge comes.

In order to get back the predicted value for the original time series data, I must add the Seasonal and Trend components with the predicted values of the Error components for those 25 time points. I can get the seasonal component since it follows a perfect cyclic pattern. But How can I get the Trend-values for those 25 time points???

Any help on this will be much appreciated. Thanks.