During the last three and a half decades various load forecasting methods have been created. However, these methods cannot forecast future loads in case of severe lack of data, when only few loads are measured in nonconsecutive years. This paper presents two simple methods developed specially to handle such situation: a simple trending method and a method based on fuzzy modelling. Fuzzy modelling is carried out using subtractive fuzzy clustering algorithm and recursive least-squares estimator. Both load forecasting methods are illustrated by numerical example.