Boost short-term load forecasts with synthetic data from transferred latent space information

Abstract Sustainable energy systems are characterised by an increased integration of renewable sources, which magnifies the fluctuations in supply. Methods to cope with these magnified fluctuations, such as load shifting, typically require accurate short-term forecasts. Although numerous machine learning models have been developed improve forecasting (STLF), often large amounts training data. Unfortunately, data is usually not available, for example, due new users or privacy concerns. Therefore, obtaining forecasts little a major challenge. The present paper thus proposes latent space-based forecast enhancer (LSFE), method combines transfer and augmentation enhance STLF when limited. LSFE first trains generative model on source similar target before using space representation generate seed noise. Finally, we use this noise synthetic data, combine real STLF. We evaluate real-world electricity examining influence its components, analysing obtained forecasts, comparing performance benchmark models. show that Latent Space-based Forecast Enhancer generally capable improving accuracy helps successfully meet challenge limited available