*3.2. Dataset*

We considered the system hourly demand and renewable power generation data obtained from the Eskom database in our study (https://www.eskom.co.za/dataportal/, accessed on 1 October 2021). Eskom is a South African electricity public utility company established to be the Electricity Supply Commission [44]. It owns and operates a number of noteworthy power plants that provide roughly 95% of South Africa's electricity [45]. Eskom provides data on power generated, consumed, and from open cycle gas turbines, renewables and power outages. Our research focused on demand-side data and renewable energy sources, which are reflective of a typical smart grid. Thus, the use cases are as follows:


The above use cases were selected because a typical grid-tied microgrid in a smart grid system can be expected to supply power to consumers both from the main grid as well as from local renewable sources. Consequently, when in the grid-tied mode, the system hourly actual demand dataset suffices as a relevant use case for prediction purposes. On the other hand, when in an island mode, the hourly renewable generation dataset becomes relevant for forecasting sake. Thus, both use cases are useful for designing energy management systems capable of day-ahead or week-ahead demand and supply forecasting in smart grids. Such knowledge helps to anticipate the amount of electricity that will be used hourly so that sufficient generation can be made available to meet such electricity demand.
