In this paper, a number of approaches to the modelling of electricity demand, on a variety of time-scales, are considered. These approaches fall under the category of 'intelligent' systems engineering, where techniques such as neural networks, fuzzy logic and genetic algorithms are employed. The paper attempts to give some motivation for the employment of such techniques, while also making some effort to be realistic about the limitations of such methods, in particular a number of important caveats that should be borne in mind when utilising these techniques within the current application domain. In general, the electricity demand data is modelled as a time series, but one application considered involves application of linguistic modelling to capture operator expertise.