Short Term Load Forecasting for Individual Consumers based on Markov Chains

Emerging smart-grid applications in low-voltage systems generate a need to forecast not only aggregated load profiles, b ut i ndividual, c onsumer s pecific pr ofiles in a high temporal resolution. These load profiles a re n oisy a nd volatile, which makes forecasting them challenging. A short term load forecasting (STLF) model, based on Markov chains, is presented in this paper. It is designed for continuous training during operation and can be pre-trained, making it universally applicable. Its performance is evaluated at a synthetic dataset of individual households. The predictive accuracy of the model is investigated for different forecasting resolutions and levels of load aggregation. The results show that the general characteristics of individual, high resolution load profiles i s m odelled reasonably well, outperforming a naive prediction, utilizing a scaled standard load profile.