Electricity demand is a time series which consist of a long-term repeating cycle of one year (e.g. more demand in winter due to heating needed), a medium-term repeating cycle of one week (e.g. less demand on Sunday due to factory closed), a short-term repeating cycle of one day (e.g. less demand on mid-night due to less human activities), a temperature dependent variable (if the winter is colder than a normal one, electricity demand will also be higher) and a random variable which you could not explain.

Therefore, the easiest way to predict hourly electrical load (5 days ahead), you need to know the long, medium and short-term cycles and a temperature dependent variable (you can get it from analysing the historical data) and the temperature forecast from local observatory (most of them provide 7-day weather forecast).

Added to the previous answer, I would suggest the use of neural networks as tool to process the previously mentioned data. Matlab has already embedded this function, otherwise you can build your own code.

Is neural networks method a frugal type in terms of processing power? I'm looking for simple way for embeded solutions for micro-grid applications.